diff --git a/FewShot/39tasks/5shot/FromRKD-Timewar_Processingline_v2/seed0/default/train_5shot_seed0_processingline_5k/checkpoint-1000/config.json b/FewShot/39tasks/5shot/FromRKD-Timewar_Processingline_v2/seed0/default/train_5shot_seed0_processingline_5k/checkpoint-1000/config.json new file mode 100644 index 0000000000000000000000000000000000000000..4740d0455bef31c469ee5dab6d4a47d64df61665 --- /dev/null +++ b/FewShot/39tasks/5shot/FromRKD-Timewar_Processingline_v2/seed0/default/train_5shot_seed0_processingline_5k/checkpoint-1000/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": 0.5, + "rkd_loss_weight_end": 0.0, + "rkd_loss_weight_schedule": null, + "rkd_loss_weight_start": 0.1, + "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/FewShot/39tasks/5shot/FromRKD-Timewar_Processingline_v2/seed0/default/train_5shot_seed0_processingline_5k/checkpoint-1000/experiment_cfg/metadata.json b/FewShot/39tasks/5shot/FromRKD-Timewar_Processingline_v2/seed0/default/train_5shot_seed0_processingline_5k/checkpoint-1000/experiment_cfg/metadata.json new file mode 100644 index 0000000000000000000000000000000000000000..6f73f8489294c58acc82d1351a17e5e3ff316e18 --- /dev/null +++ b/FewShot/39tasks/5shot/FromRKD-Timewar_Processingline_v2/seed0/default/train_5shot_seed0_processingline_5k/checkpoint-1000/experiment_cfg/metadata.json @@ -0,0 +1,431 @@ +{ + "new_embodiment": { + "statistics": { + "state": { + "base_position": { + "max": [ + 7.778976917266846, + 0.33403685688972473, + 0.7223060727119446 + ], + "min": [ + -4.821934223175049, + -6.785110950469971, + 0.6994727253913879 + ], + "mean": [ + 2.5886094736594543, + -2.1559664577316586, + 0.7003580854548973 + ], + "std": [ + 1.6717149343420803, + 1.3743695399555862, + 0.000940069542359506 + ], + "q01": [ + -0.7517815060249613, + -5.497849509089705, + 0.6999776914424374 + ], + "q99": [ + 6.060945690807028, + -0.5331499932493624, + 0.7028864260472257 + ] + }, + "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, + 0.24488939171811902, + 0.5968549204077488 + ], + "std": [ + 0.0, + 0.0, + 0.6659970734507764, + 0.37443298392262264 + ], + "q01": [ + 0.0, + 0.0, + -0.9999999999999999, + 4.149924358336069e-06 + ], + "q99": [ + 0.0, + 0.0, + 0.9999999999999999, + 0.9999999999999999 + ] + }, + "end_effector_position_relative": { + "max": [ + 0.8848381638526917, + 0.8569459915161133, + 1.053884744644165 + ], + "min": [ + -0.5828549861907959, + -0.8486236333847046, + -0.3446122705936432 + ], + "mean": [ + 0.25087486209693133, + -0.005738920462036393, + 0.43643948866361787 + ], + "std": [ + 0.15741544532518012, + 0.2547603106793105, + 0.2050434492495558 + ], + "q01": [ + -0.013990057288082561, + -0.4824599092623036, + 0.18825220353500022 + ], + "q99": [ + 0.5411989642841469, + 0.4754923657245658, + 0.6912323747882034 + ] + }, + "end_effector_rotation_relative": { + "max": [ + 0.9999949932098389, + 0.9997354745864868, + 0.9900747537612915, + 0.9735639095306396 + ], + "min": [ + -0.9999914169311523, + -0.9999884366989136, + -0.9975369572639465, + 6.754255821306288e-08 + ], + "mean": [ + -0.22290331318740714, + 0.04579037037503981, + -0.05804094906527823, + 0.12347219024917105 + ], + "std": [ + 0.8519267931761344, + 0.2949394732270141, + 0.30026078832556913, + 0.16333086331829588 + ], + "q01": [ + -0.9957959450418518, + -0.5678869493126506, + -0.5318135475761875, + 0.0014160465686576505 + ], + "q99": [ + 0.9900692296898317, + 0.6357358195673684, + 0.5009847868015971, + 0.4287044482210845 + ] + }, + "gripper_qpos": { + "max": [ + 0.05289580300450325, + 0.00937709677964449 + ], + "min": [ + -0.011669117026031017, + -0.06307756155729294 + ], + "mean": [ + 0.03203268930378034, + -0.032331639271968075 + ], + "std": [ + 0.011607049173153697, + 0.011306286745872529 + ], + "q01": [ + 0.006429109393656672, + -0.04056586943958638 + ], + "q99": [ + 0.040377860142451545, + -0.006593889698673258 + ] + } + }, + "action": { + "base_motion": { + "max": [ + 1.0, + 1.0, + 1.0, + 0.0 + ], + "min": [ + -1.0, + -1.0, + -1.0, + 0.0 + ], + "mean": [ + 0.006050336852760097, + 0.002728916466628117, + -0.0017236811393598838, + 0.0 + ], + "std": [ + 0.11032167711449951, + 0.09824140301615687, + 0.07825208047515658, + 0.0 + ], + "q01": [ + -0.154721262094005, + -0.18955302988697728, + -0.14454065867754942, + 0.0 + ], + "q99": [ + 0.24492955247744197, + 0.2496797822190429, + 0.11966742325336034, + 0.0 + ] + }, + "control_mode": { + "max": [ + 1.0 + ], + "min": [ + -1.0 + ], + "mean": [ + -0.9294448715752568 + ], + "std": [ + 0.36897903503358326 + ], + "q01": [ + -1.0 + ], + "q99": [ + -0.21627562513311385 + ] + }, + "end_effector_position": { + "max": [ + 1.0, + 1.0, + 1.0 + ], + "min": [ + -1.0, + -1.0, + -1.0 + ], + "mean": [ + -0.00826780165926883, + 0.003161997359379232, + -0.08157453651841007 + ], + "std": [ + 0.4545430944464769, + 0.41040151434986594, + 0.4102644251536544 + ], + "q01": [ + -0.9653185255999107, + -0.9255635199318754, + -0.8208437313341718 + ], + "q99": [ + 0.7987781084592441, + 0.9260844232330088, + 0.8357421812622201 + ] + }, + "end_effector_rotation": { + "max": [ + 1.0, + 1.0, + 1.0 + ], + "min": [ + -1.0, + -1.0, + -1.0 + ], + "mean": [ + 0.007367024641295137, + -0.02060756343728361, + -0.001884833075998489 + ], + "std": [ + 0.1116768956941076, + 0.13249042272165212, + 0.12338908364725322 + ], + "q01": [ + -0.26894658663246845, + -0.39865607740339326, + -0.34393735191302655 + ], + "q99": [ + 0.3191674780171587, + 0.31133630397431267, + 0.33968476423051097 + ] + }, + "gripper_close": { + "max": [ + 1.0 + ], + "min": [ + -1.0 + ], + "mean": [ + -0.3098994894722959 + ], + "std": [ + 0.9507718287498876 + ], + "q01": [ + -1.0 + ], + "q99": [ + 0.7545620899936146 + ] + } + } + }, + "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/FewShot/39tasks/5shot/FromRKD-Timewar_Processingline_v2/seed0/default/train_5shot_seed0_processingline_5k/checkpoint-1000/model.safetensors.index.json b/FewShot/39tasks/5shot/FromRKD-Timewar_Processingline_v2/seed0/default/train_5shot_seed0_processingline_5k/checkpoint-1000/model.safetensors.index.json new file mode 100644 index 0000000000000000000000000000000000000000..b20baed71f4d16a411223b70d45d934c6cbbfe97 --- /dev/null +++ b/FewShot/39tasks/5shot/FromRKD-Timewar_Processingline_v2/seed0/default/train_5shot_seed0_processingline_5k/checkpoint-1000/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/FewShot/39tasks/5shot/FromRKD-Timewar_Processingline_v2/seed0/default/train_5shot_seed0_processingline_5k/checkpoint-1000/trainer_state.json b/FewShot/39tasks/5shot/FromRKD-Timewar_Processingline_v2/seed0/default/train_5shot_seed0_processingline_5k/checkpoint-1000/trainer_state.json new file mode 100644 index 0000000000000000000000000000000000000000..fb451d5e51177e9c7b6a4ca427fadcf771ffbe7e --- /dev/null +++ b/FewShot/39tasks/5shot/FromRKD-Timewar_Processingline_v2/seed0/default/train_5shot_seed0_processingline_5k/checkpoint-1000/trainer_state.json @@ -0,0 +1,4634 @@ +{ + "best_global_step": null, + "best_metric": null, + "best_model_checkpoint": null, + "epoch": 1.6778523489932886, + "eval_steps": 500, + "global_step": 1000, + "is_hyper_param_search": false, + "is_local_process_zero": true, + "is_world_process_zero": true, + "log_history": [ + { + "epoch": 0.016778523489932886, + "flow_matching_loss": 0.2056944653391838, + "grad_norm": 0.9650156497955322, + "learning_rate": 1.08e-06, + "loss": 0.3957, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.13596224263310433, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.10807111635804176, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.37999560236930846, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.2056944653391838, + "weighted_rkd_loss": 0.18999780118465423 + }, + { + "epoch": 0.03355704697986577, + "flow_matching_loss": 0.24405399858951568, + "grad_norm": 1.1819026470184326, + "learning_rate": 2.28e-06, + "loss": 0.4373, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.13781716972589492, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.11078641712665557, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.3864207535982132, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.24405399858951568, + "weighted_rkd_loss": 0.1932103767991066 + }, + { + "epoch": 0.050335570469798654, + "flow_matching_loss": 0.18607925474643708, + "grad_norm": 1.1378169059753418, + "learning_rate": 3.48e-06, + "loss": 0.3642, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.12805165946483613, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.10023756325244904, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.3563408851623535, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.18607925474643708, + "weighted_rkd_loss": 0.17817044258117676 + }, + { + "epoch": 0.06711409395973154, + "flow_matching_loss": 0.1954496458172798, + "grad_norm": 0.9604244232177734, + "learning_rate": 4.68e-06, + "loss": 0.3626, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.12190236821770668, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.09053741469979286, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.33434214890003205, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.1954496458172798, + "weighted_rkd_loss": 0.16717107445001603 + }, + { + "epoch": 0.08389261744966443, + "flow_matching_loss": 0.18862971812486648, + "grad_norm": 1.490982174873352, + "learning_rate": 5.8800000000000005e-06, + "loss": 0.3556, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.12135515362024307, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.09131846278905868, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.33402876555919647, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.18862971812486648, + "weighted_rkd_loss": 0.16701438277959824 + }, + { + "epoch": 0.10067114093959731, + "flow_matching_loss": 0.1686223641037941, + "grad_norm": 0.6995788812637329, + "learning_rate": 7.08e-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": 0.11892070844769478, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.08481515794992447, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.322656574845314, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.1686223641037941, + "weighted_rkd_loss": 0.161328287422657 + }, + { + "epoch": 0.1174496644295302, + "flow_matching_loss": 0.1650075815618038, + "grad_norm": 0.7736762762069702, + "learning_rate": 8.28e-06, + "loss": 0.3138, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.10964272096753121, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.07826093062758446, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.2975463718175888, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.1650075815618038, + "weighted_rkd_loss": 0.1487731859087944 + }, + { + "epoch": 0.1342281879194631, + "flow_matching_loss": 0.14050063565373422, + "grad_norm": 0.7896946668624878, + "learning_rate": 9.48e-06, + "loss": 0.2771, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.10285875871777535, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.06756566911935806, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.27328317910432814, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.14050063565373422, + "weighted_rkd_loss": 0.13664158955216407 + }, + { + "epoch": 0.15100671140939598, + "flow_matching_loss": 0.12970626056194307, + "grad_norm": 0.6709607839584351, + "learning_rate": 1.068e-05, + "loss": 0.2633, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.09869295060634613, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.06970738619565964, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.267093288898468, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.12970626056194307, + "weighted_rkd_loss": 0.133546644449234 + }, + { + "epoch": 0.16778523489932887, + "flow_matching_loss": 0.11140652000904083, + "grad_norm": 0.6956789493560791, + "learning_rate": 1.1880000000000001e-05, + "loss": 0.2261, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.0861569106578827, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.05709078311920166, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.22940460294485093, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.11140652000904083, + "weighted_rkd_loss": 0.11470230147242547 + }, + { + "epoch": 0.18456375838926176, + "flow_matching_loss": 0.09781593903899193, + "grad_norm": 0.7636014223098755, + "learning_rate": 1.308e-05, + "loss": 0.2177, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.09013360589742661, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.059417485073208806, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.23968469649553298, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.09781593903899193, + "weighted_rkd_loss": 0.11984234824776649 + }, + { + "epoch": 0.20134228187919462, + "flow_matching_loss": 0.08636336326599121, + "grad_norm": 0.6833075881004333, + "learning_rate": 1.428e-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": 0.0791482873260975, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.052746016159653665, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.2110425904393196, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.08636336326599121, + "weighted_rkd_loss": 0.1055212952196598 + }, + { + "epoch": 0.2181208053691275, + "flow_matching_loss": 0.08181707337498664, + "grad_norm": 0.5722662210464478, + "learning_rate": 1.548e-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": 0.0808029130101204, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.05019080862402916, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.2117966368794441, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.08181707337498664, + "weighted_rkd_loss": 0.10589831843972206 + }, + { + "epoch": 0.2348993288590604, + "flow_matching_loss": 0.08008071929216384, + "grad_norm": 0.5205653309822083, + "learning_rate": 1.6680000000000003e-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": 0.07563451901078225, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.046241700649261475, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.19751073718070983, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.08008071929216384, + "weighted_rkd_loss": 0.09875536859035491 + }, + { + "epoch": 0.2516778523489933, + "flow_matching_loss": 0.07708811983466149, + "grad_norm": 0.6222124099731445, + "learning_rate": 1.7879999999999998e-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": 0.07161490805447102, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.04470515791326761, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.18793497681617738, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.07708811983466149, + "weighted_rkd_loss": 0.09396748840808869 + }, + { + "epoch": 0.2684563758389262, + "flow_matching_loss": 0.07064299248158931, + "grad_norm": 0.5130728483200073, + "learning_rate": 1.908e-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": 0.06377769745886326, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.038518281653523445, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.166073676943779, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.07064299248158931, + "weighted_rkd_loss": 0.0830368384718895 + }, + { + "epoch": 0.28523489932885904, + "flow_matching_loss": 0.0708860356360674, + "grad_norm": 0.44459694623947144, + "learning_rate": 2.0280000000000002e-05, + "loss": 0.1507, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.06139358542859554, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.03676006477326155, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.15954723656177522, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.0708860356360674, + "weighted_rkd_loss": 0.07977361828088761 + }, + { + "epoch": 0.30201342281879195, + "flow_matching_loss": 0.06621370911598205, + "grad_norm": 0.43706104159355164, + "learning_rate": 2.148e-05, + "loss": 0.1474, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.062162948027253154, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.03814310748130083, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.16246900260448455, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.06621370911598205, + "weighted_rkd_loss": 0.08123450130224227 + }, + { + "epoch": 0.3187919463087248, + "flow_matching_loss": 0.06586643978953362, + "grad_norm": 0.48638027906417847, + "learning_rate": 2.268e-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": 0.058573270216584206, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.03490185085684061, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.15204839110374452, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.06586643978953362, + "weighted_rkd_loss": 0.07602419555187226 + }, + { + "epoch": 0.33557046979865773, + "flow_matching_loss": 0.06912102214992047, + "grad_norm": 0.45584234595298767, + "learning_rate": 2.3880000000000002e-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": 0.0558802418410778, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.03365606628358364, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.1454165495932102, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.06912102214992047, + "weighted_rkd_loss": 0.0727082747966051 + }, + { + "epoch": 0.3523489932885906, + "flow_matching_loss": 0.06224445998668671, + "grad_norm": 0.516363799571991, + "learning_rate": 2.508e-05, + "loss": 0.1261, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.04999670423567295, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.027720438688993453, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.12771384567022323, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.06224445998668671, + "weighted_rkd_loss": 0.06385692283511162 + }, + { + "epoch": 0.3691275167785235, + "flow_matching_loss": 0.06367993578314782, + "grad_norm": 0.4426209032535553, + "learning_rate": 2.628e-05, + "loss": 0.1292, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.050844109430909154, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.02938362695276737, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.1310718484222889, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.06367993578314782, + "weighted_rkd_loss": 0.06553592421114444 + }, + { + "epoch": 0.3859060402684564, + "flow_matching_loss": 0.061497534066438674, + "grad_norm": 0.4488156735897064, + "learning_rate": 2.748e-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": 0.04617831408977509, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.02737251725047827, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.11972914338111877, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.061497534066438674, + "weighted_rkd_loss": 0.059864571690559386 + }, + { + "epoch": 0.40268456375838924, + "flow_matching_loss": 0.05670501254498959, + "grad_norm": 0.4305253028869629, + "learning_rate": 2.868e-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": 0.04511659294366836, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.02554646022617817, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.11577964574098587, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.05670501254498959, + "weighted_rkd_loss": 0.057889822870492935 + }, + { + "epoch": 0.41946308724832215, + "flow_matching_loss": 0.052361448854207994, + "grad_norm": 0.39369747042655945, + "learning_rate": 2.9880000000000002e-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": 0.04291637688875198, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.021558509021997452, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.10739126279950142, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.052361448854207994, + "weighted_rkd_loss": 0.05369563139975071 + }, + { + "epoch": 0.436241610738255, + "flow_matching_loss": 0.05451476089656353, + "grad_norm": 0.5052051544189453, + "learning_rate": 2.9999734259635676e-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": 0.04195186048746109, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.02312448900192976, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.10702821090817452, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.05451476089656353, + "weighted_rkd_loss": 0.05351410545408726 + }, + { + "epoch": 0.45302013422818793, + "flow_matching_loss": 0.060980511084198954, + "grad_norm": 0.4307898283004761, + "learning_rate": 2.9998815663057245e-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": 0.04224815517663956, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.023023180663585663, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.10751949027180671, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.060980511084198954, + "weighted_rkd_loss": 0.053759745135903356 + }, + { + "epoch": 0.4697986577181208, + "flow_matching_loss": 0.050894477590918544, + "grad_norm": 0.395856112241745, + "learning_rate": 2.999724096969228e-05, + "loss": 0.1016, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.039661315828561784, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.022144765593111515, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.10146739631891251, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.050894477590918544, + "weighted_rkd_loss": 0.050733698159456254 + }, + { + "epoch": 0.4865771812080537, + "flow_matching_loss": 0.05287523865699768, + "grad_norm": 0.3969660699367523, + "learning_rate": 2.9995010248422984e-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": 0.03771056812256575, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.019884974136948584, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.09530610963702202, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.05287523865699768, + "weighted_rkd_loss": 0.04765305481851101 + }, + { + "epoch": 0.5033557046979866, + "flow_matching_loss": 0.0479743555188179, + "grad_norm": 0.36045897006988525, + "learning_rate": 2.999212359682832e-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": 0.03238320779055357, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.016917126439511775, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.08168354257941246, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.0479743555188179, + "weighted_rkd_loss": 0.04084177128970623 + }, + { + "epoch": 0.5201342281879194, + "flow_matching_loss": 0.04859849065542221, + "grad_norm": 0.39492926001548767, + "learning_rate": 2.9988581141179787e-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": 0.034291915595531464, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.016848549619317053, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.08543238118290901, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.04859849065542221, + "weighted_rkd_loss": 0.042716190591454505 + }, + { + "epoch": 0.5369127516778524, + "flow_matching_loss": 0.04939671419560909, + "grad_norm": 0.4409841299057007, + "learning_rate": 2.998438303643587e-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": 0.03414266463369131, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.018428620789200066, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.08671394959092141, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.04939671419560909, + "weighted_rkd_loss": 0.043356974795460704 + }, + { + "epoch": 0.5536912751677853, + "flow_matching_loss": 0.04955456294119358, + "grad_norm": 0.37391889095306396, + "learning_rate": 2.997952946623527e-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": 0.033923293277621266, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.018506798055022956, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.0863533854484558, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.04955456294119358, + "weighted_rkd_loss": 0.0431766927242279 + }, + { + "epoch": 0.5704697986577181, + "flow_matching_loss": 0.050061154365539554, + "grad_norm": 0.40724676847457886, + "learning_rate": 2.9974020642888875e-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": 0.03251873888075352, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.018702817521989344, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.08374029472470283, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.050061154365539554, + "weighted_rkd_loss": 0.041870147362351416 + }, + { + "epoch": 0.587248322147651, + "flow_matching_loss": 0.04783603325486183, + "grad_norm": 0.3322409391403198, + "learning_rate": 2.9967856807370458e-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": 0.02946928795427084, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.016537870932370424, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.07547644712030888, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.04783603325486183, + "weighted_rkd_loss": 0.03773822356015444 + }, + { + "epoch": 0.6040268456375839, + "flow_matching_loss": 0.04893294684588909, + "grad_norm": 0.37779635190963745, + "learning_rate": 2.996103822930615e-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": 0.03286735378205776, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.01721991701051593, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.08295462429523467, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.04893294684588909, + "weighted_rkd_loss": 0.04147731214761734 + }, + { + "epoch": 0.6208053691275168, + "flow_matching_loss": 0.04373093023896217, + "grad_norm": 0.3719472587108612, + "learning_rate": 2.9953565206962653e-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": 0.030151859484612943, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.015161976683884859, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.07546569555997848, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.04373093023896217, + "weighted_rkd_loss": 0.03773284777998924 + }, + { + "epoch": 0.6375838926174496, + "flow_matching_loss": 0.0451244730502367, + "grad_norm": 0.41764843463897705, + "learning_rate": 2.9945438067234172e-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": 0.031219374760985373, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.016777731571346523, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.07921648174524307, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.0451244730502367, + "weighted_rkd_loss": 0.039608240872621533 + }, + { + "epoch": 0.6543624161073825, + "flow_matching_loss": 0.04234543144702911, + "grad_norm": 0.3251534402370453, + "learning_rate": 2.9936657165628124e-05, + "loss": 0.0805, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.03002563174813986, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.01633982127532363, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.07639108411967754, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.04234543144702911, + "weighted_rkd_loss": 0.03819554205983877 + }, + { + "epoch": 0.6711409395973155, + "flow_matching_loss": 0.04180084690451622, + "grad_norm": 0.3378938138484955, + "learning_rate": 2.992722288624959e-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": 0.026612223125994205, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.014572171401232481, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.06779661700129509, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.04180084690451622, + "weighted_rkd_loss": 0.033898308500647546 + }, + { + "epoch": 0.6879194630872483, + "flow_matching_loss": 0.04224371202290058, + "grad_norm": 0.3396628797054291, + "learning_rate": 2.9917135641784515e-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": 0.025039970688521863, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0127363421022892, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.06281628347933292, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.04224371202290058, + "weighted_rkd_loss": 0.03140814173966646 + }, + { + "epoch": 0.7046979865771812, + "flow_matching_loss": 0.03971978519111872, + "grad_norm": 0.30670905113220215, + "learning_rate": 2.9906395873481655e-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": 0.02627517990767956, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.013494232017546893, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.06604459211230278, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.03971978519111872, + "weighted_rkd_loss": 0.03302229605615139 + }, + { + "epoch": 0.7214765100671141, + "flow_matching_loss": 0.04066108614206314, + "grad_norm": 0.39300358295440674, + "learning_rate": 2.989500405113326e-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": 0.026304544880986214, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.01363995224237442, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.06624904237687587, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.04066108614206314, + "weighted_rkd_loss": 0.033124521188437936 + }, + { + "epoch": 0.738255033557047, + "flow_matching_loss": 0.04158542156219482, + "grad_norm": 0.3305646777153015, + "learning_rate": 2.9882960673054547e-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": 0.025151407718658446, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.01327982721850276, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.06358264274895191, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.04158542156219482, + "weighted_rkd_loss": 0.031791321374475955 + }, + { + "epoch": 0.7550335570469798, + "flow_matching_loss": 0.04144757613539696, + "grad_norm": 0.34595179557800293, + "learning_rate": 2.9870266266061886e-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": 0.022307384945452213, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0116576106287539, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.0562723807990551, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.04144757613539696, + "weighted_rkd_loss": 0.02813619039952755 + }, + { + "epoch": 0.7718120805369127, + "flow_matching_loss": 0.04313542582094669, + "grad_norm": 0.3158428370952606, + "learning_rate": 2.9856921385449766e-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": 0.02726253420114517, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.014133189246058464, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.0686582576483488, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.04313542582094669, + "weighted_rkd_loss": 0.0343291288241744 + }, + { + "epoch": 0.7885906040268457, + "flow_matching_loss": 0.042977604642510416, + "grad_norm": 0.3631656765937805, + "learning_rate": 2.9842926614966502e-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": 0.023044148832559584, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.012475840467959642, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.05856413841247558, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.042977604642510416, + "weighted_rkd_loss": 0.02928206920623779 + }, + { + "epoch": 0.8053691275167785, + "flow_matching_loss": 0.03909936882555485, + "grad_norm": 0.29421156644821167, + "learning_rate": 2.982828256678868e-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": 0.02391676288098097, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0120180937461555, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.059851620346307755, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.03909936882555485, + "weighted_rkd_loss": 0.029925810173153877 + }, + { + "epoch": 0.8221476510067114, + "flow_matching_loss": 0.03855758477002382, + "grad_norm": 0.2879815399646759, + "learning_rate": 2.9812989881494412e-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": 0.02446096483618021, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.012417732179164887, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.061339662224054334, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.03855758477002382, + "weighted_rkd_loss": 0.030669831112027167 + }, + { + "epoch": 0.8389261744966443, + "flow_matching_loss": 0.04289716742932796, + "grad_norm": 0.37124210596084595, + "learning_rate": 2.9797049228035296e-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": 0.0244005236774683, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.013081625755876302, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.06188267394900322, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.04289716742932796, + "weighted_rkd_loss": 0.03094133697450161 + }, + { + "epoch": 0.8557046979865772, + "flow_matching_loss": 0.03983708620071411, + "grad_norm": 0.3242025375366211, + "learning_rate": 2.978046130370715e-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": 0.022115133143961428, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.011383708892390132, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.05561397559940815, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.03983708620071411, + "weighted_rkd_loss": 0.027806987799704076 + }, + { + "epoch": 0.87248322147651, + "flow_matching_loss": 0.03861053548753261, + "grad_norm": 0.30338215827941895, + "learning_rate": 2.9763226834119535e-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": 0.021896894834935666, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.011009396938607096, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.054803187027573586, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.03861053548753261, + "weighted_rkd_loss": 0.027401593513786793 + }, + { + "epoch": 0.889261744966443, + "flow_matching_loss": 0.037504769116640094, + "grad_norm": 0.2798364460468292, + "learning_rate": 2.9745346573163963e-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": 0.019946480728685857, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.009748974908143282, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.04964193664491177, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.037504769116640094, + "weighted_rkd_loss": 0.024820968322455884 + }, + { + "epoch": 0.9060402684563759, + "flow_matching_loss": 0.03797667548060417, + "grad_norm": 0.25551819801330566, + "learning_rate": 2.9726821302980987e-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": 0.021121531259268522, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.01155223953537643, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.05379530191421509, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.03797667548060417, + "weighted_rkd_loss": 0.026897650957107545 + }, + { + "epoch": 0.9228187919463087, + "flow_matching_loss": 0.037000850029289725, + "grad_norm": 0.31982243061065674, + "learning_rate": 2.9707651833925935e-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": 0.0204544547945261, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.009314438654109835, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.05022334791719914, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.037000850029289725, + "weighted_rkd_loss": 0.02511167395859957 + }, + { + "epoch": 0.9395973154362416, + "flow_matching_loss": 0.038409977406263354, + "grad_norm": 0.32383039593696594, + "learning_rate": 2.968783900453348e-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": 0.01836938913911581, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.010035208612680434, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.046773987263441084, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.038409977406263354, + "weighted_rkd_loss": 0.023386993631720542 + }, + { + "epoch": 0.9563758389261745, + "flow_matching_loss": 0.03884415850043297, + "grad_norm": 0.3668201267719269, + "learning_rate": 2.9667383681480968e-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": 0.01754778316244483, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.008952949335798622, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.04404851570725441, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.03884415850043297, + "weighted_rkd_loss": 0.022024257853627206 + }, + { + "epoch": 0.9731543624161074, + "flow_matching_loss": 0.03858929816633463, + "grad_norm": 0.31969910860061646, + "learning_rate": 2.9646286759550493e-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": 0.018698493484407663, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.00886254720389843, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.04625953398644924, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.03858929816633463, + "weighted_rkd_loss": 0.02312976699322462 + }, + { + "epoch": 0.9899328859060402, + "flow_matching_loss": 0.037984182313084605, + "grad_norm": 0.37169548869132996, + "learning_rate": 2.9624549161589764e-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": 0.020831290166825056, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.011616774275898934, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.0532793540507555, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.037984182313084605, + "weighted_rkd_loss": 0.02663967702537775 + }, + { + "epoch": 1.0067114093959733, + "flow_matching_loss": 0.032401621155440805, + "grad_norm": 0.2810411751270294, + "learning_rate": 2.9602171838471745e-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": 0.02296459171921015, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.010374387260526419, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.05630357041954994, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.032401621155440805, + "weighted_rkd_loss": 0.02815178520977497 + }, + { + "epoch": 1.023489932885906, + "flow_matching_loss": 0.03286149390041828, + "grad_norm": 0.2521582245826721, + "learning_rate": 2.957915576905303e-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": 0.017456885334104298, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.008559525944292545, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.04347329661250114, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.03286149390041828, + "weighted_rkd_loss": 0.02173664830625057 + }, + { + "epoch": 1.0402684563758389, + "flow_matching_loss": 0.03205829132348299, + "grad_norm": 0.3185127377510071, + "learning_rate": 2.9555501960131058e-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": 0.01915504578500986, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.009150991961359978, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.04746108390390873, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.03205829132348299, + "weighted_rkd_loss": 0.023730541951954363 + }, + { + "epoch": 1.0570469798657718, + "flow_matching_loss": 0.03440665490925312, + "grad_norm": 0.2903192341327667, + "learning_rate": 2.9531211446400068e-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": 0.018206587247550487, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.009964327700436116, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.04637750200927258, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.03440665490925312, + "weighted_rkd_loss": 0.02318875100463629 + }, + { + "epoch": 1.0738255033557047, + "flow_matching_loss": 0.0327774690464139, + "grad_norm": 0.2796550691127777, + "learning_rate": 2.9506285290405813e-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": 0.01788708958774805, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.009063597302883863, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.044837776198983194, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.0327774690464139, + "weighted_rkd_loss": 0.022418888099491597 + }, + { + "epoch": 1.0906040268456376, + "flow_matching_loss": 0.03253862913697958, + "grad_norm": 0.29687777161598206, + "learning_rate": 2.948072458249911e-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": 0.018407645542174576, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.008646867936477065, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.04546215943992138, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.03253862913697958, + "weighted_rkd_loss": 0.02273107971996069 + }, + { + "epoch": 1.1073825503355705, + "flow_matching_loss": 0.034775618463754654, + "grad_norm": 0.26023975014686584, + "learning_rate": 2.9454530440788125e-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": 0.017261851113289595, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.008692631544545294, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.04321633391082287, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.034775618463754654, + "weighted_rkd_loss": 0.021608166955411436 + }, + { + "epoch": 1.1241610738255035, + "flow_matching_loss": 0.03697704430669546, + "grad_norm": 0.2973390817642212, + "learning_rate": 2.9427704011089474e-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": 0.017315912712365388, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.008650409011170267, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.043282233737409116, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.03697704430669546, + "weighted_rkd_loss": 0.021641116868704558 + }, + { + "epoch": 1.1409395973154361, + "flow_matching_loss": 0.034938321262598035, + "grad_norm": 0.33880263566970825, + "learning_rate": 2.94002464668781e-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": 0.01774648055434227, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.008731141500175, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.04422410242259502, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.034938321262598035, + "weighted_rkd_loss": 0.02211205121129751 + }, + { + "epoch": 1.157718120805369, + "flow_matching_loss": 0.03397301379591226, + "grad_norm": 0.30417579412460327, + "learning_rate": 2.9372159009235936e-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": 0.016883174236863852, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0084213403519243, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.04218768924474716, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.03397301379591226, + "weighted_rkd_loss": 0.02109384462237358 + }, + { + "epoch": 1.174496644295302, + "flow_matching_loss": 0.033245801366865634, + "grad_norm": 0.2966938614845276, + "learning_rate": 2.9343442866799365e-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": 0.01730244169011712, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.008166278852149845, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.04277116190642118, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.033245801366865634, + "weighted_rkd_loss": 0.02138558095321059 + }, + { + "epoch": 1.191275167785235, + "flow_matching_loss": 0.032241361774504186, + "grad_norm": 0.3242097496986389, + "learning_rate": 2.9314099295705495e-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": 0.01751971701160073, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.008714311430230737, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.04375374540686607, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.032241361774504186, + "weighted_rkd_loss": 0.021876872703433036 + }, + { + "epoch": 1.2080536912751678, + "flow_matching_loss": 0.030840910412371157, + "grad_norm": 0.3416939675807953, + "learning_rate": 2.928412957953718e-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": 0.016577553190290927, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.008127822168171405, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.041282928362488744, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.030840910412371157, + "weighted_rkd_loss": 0.020641464181244372 + }, + { + "epoch": 1.2248322147651007, + "flow_matching_loss": 0.030589586496353148, + "grad_norm": 0.3434118926525116, + "learning_rate": 2.925353502926689e-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": 0.014619570132344962, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.00703063509427011, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.03626977521926165, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.030589586496353148, + "weighted_rkd_loss": 0.018134887609630824 + }, + { + "epoch": 1.2416107382550337, + "flow_matching_loss": 0.028109255991876126, + "grad_norm": 0.2770143449306488, + "learning_rate": 2.922231698319937e-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": 0.015383913926780225, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.007206043181940913, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.03797387108206749, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.028109255991876126, + "weighted_rkd_loss": 0.018986935541033746 + }, + { + "epoch": 1.2583892617449663, + "flow_matching_loss": 0.03261192515492439, + "grad_norm": 0.271056592464447, + "learning_rate": 2.9190476806913088e-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": 0.016976217925548553, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.007505776965990663, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.04145821258425712, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.03261192515492439, + "weighted_rkd_loss": 0.02072910629212856 + }, + { + "epoch": 1.2751677852348993, + "flow_matching_loss": 0.030070590786635876, + "grad_norm": 0.2317705899477005, + "learning_rate": 2.9158015893200504e-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": 0.016096030082553624, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.007664719317108393, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.03985677920281887, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.030070590786635876, + "weighted_rkd_loss": 0.019928389601409434 + }, + { + "epoch": 1.2919463087248322, + "flow_matching_loss": 0.033119112625718115, + "grad_norm": 0.2506316006183624, + "learning_rate": 2.9124935662007137e-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": 0.01613023495301604, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.007737922063097358, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.03999839182943106, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.033119112625718115, + "weighted_rkd_loss": 0.01999919591471553 + }, + { + "epoch": 1.308724832214765, + "flow_matching_loss": 0.031553388386964795, + "grad_norm": 0.2889447808265686, + "learning_rate": 2.9091237560369467e-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": 0.01787243187427521, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.009260696731507779, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.04500555992126465, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.031553388386964795, + "weighted_rkd_loss": 0.022502779960632324 + }, + { + "epoch": 1.325503355704698, + "flow_matching_loss": 0.03222518637776375, + "grad_norm": 0.2785235047340393, + "learning_rate": 2.905692306235163e-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": 0.015721357706934214, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.007582925446331501, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.039025640860199926, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.03222518637776375, + "weighted_rkd_loss": 0.019512820430099963 + }, + { + "epoch": 1.342281879194631, + "flow_matching_loss": 0.030334611982107164, + "grad_norm": 0.31459397077560425, + "learning_rate": 2.9021993668980925e-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": 0.01560999769717455, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.007159262569621206, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.03837925773113966, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.030334611982107164, + "weighted_rkd_loss": 0.01918962886556983 + }, + { + "epoch": 1.3590604026845639, + "flow_matching_loss": 0.03257175628095865, + "grad_norm": 0.2993733286857605, + "learning_rate": 2.8986450908182185e-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": 0.015799124166369437, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.00794396479614079, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.039542213454842566, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.03257175628095865, + "weighted_rkd_loss": 0.019771106727421283 + }, + { + "epoch": 1.3758389261744965, + "flow_matching_loss": 0.032540197111666204, + "grad_norm": 0.31480374932289124, + "learning_rate": 2.8950296334710902e-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": 0.01558954007923603, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.007458786852657795, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.03863786738365889, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.032540197111666204, + "weighted_rkd_loss": 0.019318933691829444 + }, + { + "epoch": 1.3926174496644295, + "flow_matching_loss": 0.03170314058661461, + "grad_norm": 0.2934443950653076, + "learning_rate": 2.8913531530085248e-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": 0.016132532991468905, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.007935636956244708, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.040200702473521234, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.03170314058661461, + "weighted_rkd_loss": 0.020100351236760617 + }, + { + "epoch": 1.4093959731543624, + "flow_matching_loss": 0.03200640454888344, + "grad_norm": 0.26509004831314087, + "learning_rate": 2.887615810251687e-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": 0.014211549051105976, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.006883570412173867, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.035306668281555174, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.03200640454888344, + "weighted_rkd_loss": 0.017653334140777587 + }, + { + "epoch": 1.4261744966442953, + "flow_matching_loss": 0.02827311120927334, + "grad_norm": 0.3237679600715637, + "learning_rate": 2.8838177686840577e-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": 0.017118581011891364, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.00820444175042212, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.04244160372763872, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.02827311120927334, + "weighted_rkd_loss": 0.02122080186381936 + }, + { + "epoch": 1.4429530201342282, + "flow_matching_loss": 0.03121668491512537, + "grad_norm": 0.3042275607585907, + "learning_rate": 2.8799591944442776e-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": 0.016424584202468395, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.007744280295446515, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.040593449398875235, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.03121668491512537, + "weighted_rkd_loss": 0.020296724699437618 + }, + { + "epoch": 1.4597315436241611, + "flow_matching_loss": 0.030370122380554677, + "grad_norm": 0.3472956120967865, + "learning_rate": 2.8760402563188848e-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": 0.017987641133368017, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.009500452037900686, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.045475734397768974, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.030370122380554677, + "weighted_rkd_loss": 0.022737867198884487 + }, + { + "epoch": 1.476510067114094, + "flow_matching_loss": 0.02861998286098242, + "grad_norm": 0.24505336582660675, + "learning_rate": 2.8720611257349285e-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": 0.0141552341170609, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.006439156364649534, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.03474962469190359, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.02861998286098242, + "weighted_rkd_loss": 0.017374812345951795 + }, + { + "epoch": 1.4932885906040267, + "flow_matching_loss": 0.03184323962777853, + "grad_norm": 0.2608340084552765, + "learning_rate": 2.868021976752471e-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": 0.01494277287274599, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.00734082362614572, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.037226369604468346, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.03184323962777853, + "weighted_rkd_loss": 0.018613184802234173 + }, + { + "epoch": 1.5100671140939599, + "flow_matching_loss": 0.03042609766125679, + "grad_norm": 0.333564430475235, + "learning_rate": 2.8639229860569738e-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": 0.0160973247140646, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.007541899429634213, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.03973654881119728, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.03042609766125679, + "weighted_rkd_loss": 0.01986827440559864 + }, + { + "epoch": 1.5268456375838926, + "flow_matching_loss": 0.03012043833732605, + "grad_norm": 0.24364671111106873, + "learning_rate": 2.8597643329515688e-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": 0.014172045607119798, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.006650603329762817, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.034994694776833056, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.03012043833732605, + "weighted_rkd_loss": 0.017497347388416528 + }, + { + "epoch": 1.5436241610738255, + "flow_matching_loss": 0.03092307262122631, + "grad_norm": 0.27232593297958374, + "learning_rate": 2.855546199349214e-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": 0.014154891762882472, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.007071852730587125, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.03538163620978594, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.03092307262122631, + "weighted_rkd_loss": 0.01769081810489297 + }, + { + "epoch": 1.5604026845637584, + "flow_matching_loss": 0.027929746173322202, + "grad_norm": 0.33788374066352844, + "learning_rate": 2.85126876976474e-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": 0.012748553790152073, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.005962743982672691, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.031459851376712325, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.027929746173322202, + "weighted_rkd_loss": 0.015729925688356163 + }, + { + "epoch": 1.5771812080536913, + "flow_matching_loss": 0.028270901925861836, + "grad_norm": 0.21897122263908386, + "learning_rate": 2.8469322313067726e-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": 0.015070199593901634, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.007116558589041233, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.0372569577768445, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.028270901925861836, + "weighted_rkd_loss": 0.01862847888842225 + }, + { + "epoch": 1.5939597315436242, + "flow_matching_loss": 0.02921714950352907, + "grad_norm": 0.3024495542049408, + "learning_rate": 2.8425367736695522e-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": 0.011751286219805479, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.005542754894122481, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.02904532738029957, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.02921714950352907, + "weighted_rkd_loss": 0.014522663690149784 + }, + { + "epoch": 1.610738255033557, + "flow_matching_loss": 0.02695902418345213, + "grad_norm": 0.2782779335975647, + "learning_rate": 2.838082589124636e-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": 0.015185461472719907, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.007375222817063332, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.03774614538997412, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.02695902418345213, + "weighted_rkd_loss": 0.01887307269498706 + }, + { + "epoch": 1.62751677852349, + "flow_matching_loss": 0.026085390895605087, + "grad_norm": 0.31670045852661133, + "learning_rate": 2.8335698725124843e-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": 0.014663305971771479, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.007240531407296658, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.03656714335083962, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.026085390895605087, + "weighted_rkd_loss": 0.01828357167541981 + }, + { + "epoch": 1.6442953020134228, + "flow_matching_loss": 0.02555117029696703, + "grad_norm": 0.2121639996767044, + "learning_rate": 2.8289988212339422e-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": 0.013465657457709312, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0062181116081774235, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.03314942680299282, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.02555117029696703, + "weighted_rkd_loss": 0.01657471340149641 + }, + { + "epoch": 1.6610738255033557, + "flow_matching_loss": 0.026020037569105627, + "grad_norm": 0.22167570888996124, + "learning_rate": 2.8243696352415997e-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": 0.01538208145648241, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.008201855747029186, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.038966018334031106, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.026020037569105627, + "weighted_rkd_loss": 0.019483009167015553 + }, + { + "epoch": 1.6778523489932886, + "flow_matching_loss": 0.029859650880098343, + "grad_norm": 0.2646207809448242, + "learning_rate": 2.8196825170310495e-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": 0.012622483912855387, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.005989065254107117, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.031234032846987246, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.029859650880098343, + "weighted_rkd_loss": 0.015617016423493623 + } + ], + "logging_steps": 10, + "max_steps": 5000, + "num_input_tokens_seen": 0, + "num_train_epochs": 9, + "save_steps": 5001, + "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": 64, + "trial_name": null, + "trial_params": null +} diff --git a/FewShot/39tasks/5shot/FromRKD-Timewar_Processingline_v2/seed0/default/train_5shot_seed0_processingline_5k/checkpoint-3000/config.json b/FewShot/39tasks/5shot/FromRKD-Timewar_Processingline_v2/seed0/default/train_5shot_seed0_processingline_5k/checkpoint-3000/config.json new file mode 100644 index 0000000000000000000000000000000000000000..4740d0455bef31c469ee5dab6d4a47d64df61665 --- /dev/null +++ b/FewShot/39tasks/5shot/FromRKD-Timewar_Processingline_v2/seed0/default/train_5shot_seed0_processingline_5k/checkpoint-3000/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": 0.5, + "rkd_loss_weight_end": 0.0, + "rkd_loss_weight_schedule": null, + "rkd_loss_weight_start": 0.1, + "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/FewShot/39tasks/5shot/FromRKD-Timewar_Processingline_v2/seed0/default/train_5shot_seed0_processingline_5k/checkpoint-3000/experiment_cfg/metadata.json b/FewShot/39tasks/5shot/FromRKD-Timewar_Processingline_v2/seed0/default/train_5shot_seed0_processingline_5k/checkpoint-3000/experiment_cfg/metadata.json new file mode 100644 index 0000000000000000000000000000000000000000..6f73f8489294c58acc82d1351a17e5e3ff316e18 --- /dev/null +++ b/FewShot/39tasks/5shot/FromRKD-Timewar_Processingline_v2/seed0/default/train_5shot_seed0_processingline_5k/checkpoint-3000/experiment_cfg/metadata.json @@ -0,0 +1,431 @@ +{ + "new_embodiment": { + "statistics": { + "state": { + "base_position": { + "max": [ + 7.778976917266846, + 0.33403685688972473, + 0.7223060727119446 + ], + "min": [ + -4.821934223175049, + -6.785110950469971, + 0.6994727253913879 + ], + "mean": [ + 2.5886094736594543, + -2.1559664577316586, + 0.7003580854548973 + ], + "std": [ + 1.6717149343420803, + 1.3743695399555862, + 0.000940069542359506 + ], + "q01": [ + -0.7517815060249613, + -5.497849509089705, + 0.6999776914424374 + ], + "q99": [ + 6.060945690807028, + -0.5331499932493624, + 0.7028864260472257 + ] + }, + "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, + 0.24488939171811902, + 0.5968549204077488 + ], + "std": [ + 0.0, + 0.0, + 0.6659970734507764, + 0.37443298392262264 + ], + "q01": [ + 0.0, + 0.0, + -0.9999999999999999, + 4.149924358336069e-06 + ], + "q99": [ + 0.0, + 0.0, + 0.9999999999999999, + 0.9999999999999999 + ] + }, + "end_effector_position_relative": { + "max": [ + 0.8848381638526917, + 0.8569459915161133, + 1.053884744644165 + ], + "min": [ + -0.5828549861907959, + -0.8486236333847046, + -0.3446122705936432 + ], + "mean": [ + 0.25087486209693133, + -0.005738920462036393, + 0.43643948866361787 + ], + "std": [ + 0.15741544532518012, + 0.2547603106793105, + 0.2050434492495558 + ], + "q01": [ + -0.013990057288082561, + -0.4824599092623036, + 0.18825220353500022 + ], + "q99": [ + 0.5411989642841469, + 0.4754923657245658, + 0.6912323747882034 + ] + }, + "end_effector_rotation_relative": { + "max": [ + 0.9999949932098389, + 0.9997354745864868, + 0.9900747537612915, + 0.9735639095306396 + ], + "min": [ + -0.9999914169311523, + -0.9999884366989136, + -0.9975369572639465, + 6.754255821306288e-08 + ], + "mean": [ + -0.22290331318740714, + 0.04579037037503981, + -0.05804094906527823, + 0.12347219024917105 + ], + "std": [ + 0.8519267931761344, + 0.2949394732270141, + 0.30026078832556913, + 0.16333086331829588 + ], + "q01": [ + -0.9957959450418518, + -0.5678869493126506, + -0.5318135475761875, + 0.0014160465686576505 + ], + "q99": [ + 0.9900692296898317, + 0.6357358195673684, + 0.5009847868015971, + 0.4287044482210845 + ] + }, + "gripper_qpos": { + "max": [ + 0.05289580300450325, + 0.00937709677964449 + ], + "min": [ + -0.011669117026031017, + -0.06307756155729294 + ], + "mean": [ + 0.03203268930378034, + -0.032331639271968075 + ], + "std": [ + 0.011607049173153697, + 0.011306286745872529 + ], + "q01": [ + 0.006429109393656672, + -0.04056586943958638 + ], + "q99": [ + 0.040377860142451545, + -0.006593889698673258 + ] + } + }, + "action": { + "base_motion": { + "max": [ + 1.0, + 1.0, + 1.0, + 0.0 + ], + "min": [ + -1.0, + -1.0, + -1.0, + 0.0 + ], + "mean": [ + 0.006050336852760097, + 0.002728916466628117, + -0.0017236811393598838, + 0.0 + ], + "std": [ + 0.11032167711449951, + 0.09824140301615687, + 0.07825208047515658, + 0.0 + ], + "q01": [ + -0.154721262094005, + -0.18955302988697728, + -0.14454065867754942, + 0.0 + ], + "q99": [ + 0.24492955247744197, + 0.2496797822190429, + 0.11966742325336034, + 0.0 + ] + }, + "control_mode": { + "max": [ + 1.0 + ], + "min": [ + -1.0 + ], + "mean": [ + -0.9294448715752568 + ], + "std": [ + 0.36897903503358326 + ], + "q01": [ + -1.0 + ], + "q99": [ + -0.21627562513311385 + ] + }, + "end_effector_position": { + "max": [ + 1.0, + 1.0, + 1.0 + ], + "min": [ + -1.0, + -1.0, + -1.0 + ], + "mean": [ + -0.00826780165926883, + 0.003161997359379232, + -0.08157453651841007 + ], + "std": [ + 0.4545430944464769, + 0.41040151434986594, + 0.4102644251536544 + ], + "q01": [ + -0.9653185255999107, + -0.9255635199318754, + -0.8208437313341718 + ], + "q99": [ + 0.7987781084592441, + 0.9260844232330088, + 0.8357421812622201 + ] + }, + "end_effector_rotation": { + "max": [ + 1.0, + 1.0, + 1.0 + ], + "min": [ + -1.0, + -1.0, + -1.0 + ], + "mean": [ + 0.007367024641295137, + -0.02060756343728361, + -0.001884833075998489 + ], + "std": [ + 0.1116768956941076, + 0.13249042272165212, + 0.12338908364725322 + ], + "q01": [ + -0.26894658663246845, + -0.39865607740339326, + -0.34393735191302655 + ], + "q99": [ + 0.3191674780171587, + 0.31133630397431267, + 0.33968476423051097 + ] + }, + "gripper_close": { + "max": [ + 1.0 + ], + "min": [ + -1.0 + ], + "mean": [ + -0.3098994894722959 + ], + "std": [ + 0.9507718287498876 + ], + "q01": [ + -1.0 + ], + "q99": [ + 0.7545620899936146 + ] + } + } + }, + "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/FewShot/39tasks/5shot/FromRKD-Timewar_Processingline_v2/seed0/default/train_5shot_seed0_processingline_5k/checkpoint-3000/model.safetensors.index.json b/FewShot/39tasks/5shot/FromRKD-Timewar_Processingline_v2/seed0/default/train_5shot_seed0_processingline_5k/checkpoint-3000/model.safetensors.index.json new file mode 100644 index 0000000000000000000000000000000000000000..b20baed71f4d16a411223b70d45d934c6cbbfe97 --- /dev/null +++ b/FewShot/39tasks/5shot/FromRKD-Timewar_Processingline_v2/seed0/default/train_5shot_seed0_processingline_5k/checkpoint-3000/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/FewShot/39tasks/5shot/FromRKD-Timewar_Processingline_v2/seed0/default/train_5shot_seed0_processingline_5k/checkpoint-3000/trainer_state.json b/FewShot/39tasks/5shot/FromRKD-Timewar_Processingline_v2/seed0/default/train_5shot_seed0_processingline_5k/checkpoint-3000/trainer_state.json new file mode 100644 index 0000000000000000000000000000000000000000..7def95f2f282f635df3bd484560a5ff5ecc11ef1 --- /dev/null +++ b/FewShot/39tasks/5shot/FromRKD-Timewar_Processingline_v2/seed0/default/train_5shot_seed0_processingline_5k/checkpoint-3000/trainer_state.json @@ -0,0 +1,13834 @@ +{ + "best_global_step": null, + "best_metric": null, + "best_model_checkpoint": null, + "epoch": 5.033557046979865, + "eval_steps": 500, + "global_step": 3000, + "is_hyper_param_search": false, + "is_local_process_zero": true, + "is_world_process_zero": true, + "log_history": [ + { + "epoch": 0.016778523489932886, + "flow_matching_loss": 0.2056944653391838, + "grad_norm": 0.9650156497955322, + "learning_rate": 1.08e-06, + "loss": 0.3957, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.13596224263310433, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.10807111635804176, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.37999560236930846, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.2056944653391838, + "weighted_rkd_loss": 0.18999780118465423 + }, + { + "epoch": 0.03355704697986577, + "flow_matching_loss": 0.24405399858951568, + "grad_norm": 1.1819026470184326, + "learning_rate": 2.28e-06, + "loss": 0.4373, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.13781716972589492, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.11078641712665557, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.3864207535982132, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.24405399858951568, + "weighted_rkd_loss": 0.1932103767991066 + }, + { + "epoch": 0.050335570469798654, + "flow_matching_loss": 0.18607925474643708, + "grad_norm": 1.1378169059753418, + "learning_rate": 3.48e-06, + "loss": 0.3642, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.12805165946483613, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.10023756325244904, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.3563408851623535, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.18607925474643708, + "weighted_rkd_loss": 0.17817044258117676 + }, + { + "epoch": 0.06711409395973154, + "flow_matching_loss": 0.1954496458172798, + "grad_norm": 0.9604244232177734, + "learning_rate": 4.68e-06, + "loss": 0.3626, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.12190236821770668, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.09053741469979286, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.33434214890003205, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.1954496458172798, + "weighted_rkd_loss": 0.16717107445001603 + }, + { + "epoch": 0.08389261744966443, + "flow_matching_loss": 0.18862971812486648, + "grad_norm": 1.490982174873352, + "learning_rate": 5.8800000000000005e-06, + "loss": 0.3556, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.12135515362024307, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.09131846278905868, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.33402876555919647, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.18862971812486648, + "weighted_rkd_loss": 0.16701438277959824 + }, + { + "epoch": 0.10067114093959731, + "flow_matching_loss": 0.1686223641037941, + "grad_norm": 0.6995788812637329, + "learning_rate": 7.08e-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": 0.11892070844769478, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.08481515794992447, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.322656574845314, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.1686223641037941, + "weighted_rkd_loss": 0.161328287422657 + }, + { + "epoch": 0.1174496644295302, + "flow_matching_loss": 0.1650075815618038, + "grad_norm": 0.7736762762069702, + "learning_rate": 8.28e-06, + "loss": 0.3138, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.10964272096753121, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.07826093062758446, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.2975463718175888, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.1650075815618038, + "weighted_rkd_loss": 0.1487731859087944 + }, + { + "epoch": 0.1342281879194631, + "flow_matching_loss": 0.14050063565373422, + "grad_norm": 0.7896946668624878, + "learning_rate": 9.48e-06, + "loss": 0.2771, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.10285875871777535, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.06756566911935806, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.27328317910432814, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.14050063565373422, + "weighted_rkd_loss": 0.13664158955216407 + }, + { + "epoch": 0.15100671140939598, + "flow_matching_loss": 0.12970626056194307, + "grad_norm": 0.6709607839584351, + "learning_rate": 1.068e-05, + "loss": 0.2633, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.09869295060634613, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.06970738619565964, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.267093288898468, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.12970626056194307, + "weighted_rkd_loss": 0.133546644449234 + }, + { + "epoch": 0.16778523489932887, + "flow_matching_loss": 0.11140652000904083, + "grad_norm": 0.6956789493560791, + "learning_rate": 1.1880000000000001e-05, + "loss": 0.2261, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.0861569106578827, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.05709078311920166, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.22940460294485093, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.11140652000904083, + "weighted_rkd_loss": 0.11470230147242547 + }, + { + "epoch": 0.18456375838926176, + "flow_matching_loss": 0.09781593903899193, + "grad_norm": 0.7636014223098755, + "learning_rate": 1.308e-05, + "loss": 0.2177, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.09013360589742661, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.059417485073208806, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.23968469649553298, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.09781593903899193, + "weighted_rkd_loss": 0.11984234824776649 + }, + { + "epoch": 0.20134228187919462, + "flow_matching_loss": 0.08636336326599121, + "grad_norm": 0.6833075881004333, + "learning_rate": 1.428e-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": 0.0791482873260975, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.052746016159653665, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.2110425904393196, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.08636336326599121, + "weighted_rkd_loss": 0.1055212952196598 + }, + { + "epoch": 0.2181208053691275, + "flow_matching_loss": 0.08181707337498664, + "grad_norm": 0.5722662210464478, + "learning_rate": 1.548e-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": 0.0808029130101204, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.05019080862402916, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.2117966368794441, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.08181707337498664, + "weighted_rkd_loss": 0.10589831843972206 + }, + { + "epoch": 0.2348993288590604, + "flow_matching_loss": 0.08008071929216384, + "grad_norm": 0.5205653309822083, + "learning_rate": 1.6680000000000003e-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": 0.07563451901078225, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.046241700649261475, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.19751073718070983, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.08008071929216384, + "weighted_rkd_loss": 0.09875536859035491 + }, + { + "epoch": 0.2516778523489933, + "flow_matching_loss": 0.07708811983466149, + "grad_norm": 0.6222124099731445, + "learning_rate": 1.7879999999999998e-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": 0.07161490805447102, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.04470515791326761, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.18793497681617738, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.07708811983466149, + "weighted_rkd_loss": 0.09396748840808869 + }, + { + "epoch": 0.2684563758389262, + "flow_matching_loss": 0.07064299248158931, + "grad_norm": 0.5130728483200073, + "learning_rate": 1.908e-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": 0.06377769745886326, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.038518281653523445, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.166073676943779, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.07064299248158931, + "weighted_rkd_loss": 0.0830368384718895 + }, + { + "epoch": 0.28523489932885904, + "flow_matching_loss": 0.0708860356360674, + "grad_norm": 0.44459694623947144, + "learning_rate": 2.0280000000000002e-05, + "loss": 0.1507, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.06139358542859554, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.03676006477326155, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.15954723656177522, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.0708860356360674, + "weighted_rkd_loss": 0.07977361828088761 + }, + { + "epoch": 0.30201342281879195, + "flow_matching_loss": 0.06621370911598205, + "grad_norm": 0.43706104159355164, + "learning_rate": 2.148e-05, + "loss": 0.1474, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.062162948027253154, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.03814310748130083, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.16246900260448455, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.06621370911598205, + "weighted_rkd_loss": 0.08123450130224227 + }, + { + "epoch": 0.3187919463087248, + "flow_matching_loss": 0.06586643978953362, + "grad_norm": 0.48638027906417847, + "learning_rate": 2.268e-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": 0.058573270216584206, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.03490185085684061, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.15204839110374452, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.06586643978953362, + "weighted_rkd_loss": 0.07602419555187226 + }, + { + "epoch": 0.33557046979865773, + "flow_matching_loss": 0.06912102214992047, + "grad_norm": 0.45584234595298767, + "learning_rate": 2.3880000000000002e-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": 0.0558802418410778, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.03365606628358364, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.1454165495932102, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.06912102214992047, + "weighted_rkd_loss": 0.0727082747966051 + }, + { + "epoch": 0.3523489932885906, + "flow_matching_loss": 0.06224445998668671, + "grad_norm": 0.516363799571991, + "learning_rate": 2.508e-05, + "loss": 0.1261, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.04999670423567295, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.027720438688993453, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.12771384567022323, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.06224445998668671, + "weighted_rkd_loss": 0.06385692283511162 + }, + { + "epoch": 0.3691275167785235, + "flow_matching_loss": 0.06367993578314782, + "grad_norm": 0.4426209032535553, + "learning_rate": 2.628e-05, + "loss": 0.1292, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.050844109430909154, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.02938362695276737, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.1310718484222889, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.06367993578314782, + "weighted_rkd_loss": 0.06553592421114444 + }, + { + "epoch": 0.3859060402684564, + "flow_matching_loss": 0.061497534066438674, + "grad_norm": 0.4488156735897064, + "learning_rate": 2.748e-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": 0.04617831408977509, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.02737251725047827, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.11972914338111877, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.061497534066438674, + "weighted_rkd_loss": 0.059864571690559386 + }, + { + "epoch": 0.40268456375838924, + "flow_matching_loss": 0.05670501254498959, + "grad_norm": 0.4305253028869629, + "learning_rate": 2.868e-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": 0.04511659294366836, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.02554646022617817, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.11577964574098587, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.05670501254498959, + "weighted_rkd_loss": 0.057889822870492935 + }, + { + "epoch": 0.41946308724832215, + "flow_matching_loss": 0.052361448854207994, + "grad_norm": 0.39369747042655945, + "learning_rate": 2.9880000000000002e-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": 0.04291637688875198, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.021558509021997452, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.10739126279950142, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.052361448854207994, + "weighted_rkd_loss": 0.05369563139975071 + }, + { + "epoch": 0.436241610738255, + "flow_matching_loss": 0.05451476089656353, + "grad_norm": 0.5052051544189453, + "learning_rate": 2.9999734259635676e-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": 0.04195186048746109, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.02312448900192976, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.10702821090817452, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.05451476089656353, + "weighted_rkd_loss": 0.05351410545408726 + }, + { + "epoch": 0.45302013422818793, + "flow_matching_loss": 0.060980511084198954, + "grad_norm": 0.4307898283004761, + "learning_rate": 2.9998815663057245e-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": 0.04224815517663956, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.023023180663585663, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.10751949027180671, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.060980511084198954, + "weighted_rkd_loss": 0.053759745135903356 + }, + { + "epoch": 0.4697986577181208, + "flow_matching_loss": 0.050894477590918544, + "grad_norm": 0.395856112241745, + "learning_rate": 2.999724096969228e-05, + "loss": 0.1016, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.039661315828561784, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.022144765593111515, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.10146739631891251, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.050894477590918544, + "weighted_rkd_loss": 0.050733698159456254 + }, + { + "epoch": 0.4865771812080537, + "flow_matching_loss": 0.05287523865699768, + "grad_norm": 0.3969660699367523, + "learning_rate": 2.9995010248422984e-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": 0.03771056812256575, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.019884974136948584, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.09530610963702202, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.05287523865699768, + "weighted_rkd_loss": 0.04765305481851101 + }, + { + "epoch": 0.5033557046979866, + "flow_matching_loss": 0.0479743555188179, + "grad_norm": 0.36045897006988525, + "learning_rate": 2.999212359682832e-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": 0.03238320779055357, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.016917126439511775, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.08168354257941246, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.0479743555188179, + "weighted_rkd_loss": 0.04084177128970623 + }, + { + "epoch": 0.5201342281879194, + "flow_matching_loss": 0.04859849065542221, + "grad_norm": 0.39492926001548767, + "learning_rate": 2.9988581141179787e-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": 0.034291915595531464, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.016848549619317053, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.08543238118290901, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.04859849065542221, + "weighted_rkd_loss": 0.042716190591454505 + }, + { + "epoch": 0.5369127516778524, + "flow_matching_loss": 0.04939671419560909, + "grad_norm": 0.4409841299057007, + "learning_rate": 2.998438303643587e-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": 0.03414266463369131, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.018428620789200066, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.08671394959092141, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.04939671419560909, + "weighted_rkd_loss": 0.043356974795460704 + }, + { + "epoch": 0.5536912751677853, + "flow_matching_loss": 0.04955456294119358, + "grad_norm": 0.37391889095306396, + "learning_rate": 2.997952946623527e-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": 0.033923293277621266, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.018506798055022956, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.0863533854484558, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.04955456294119358, + "weighted_rkd_loss": 0.0431766927242279 + }, + { + "epoch": 0.5704697986577181, + "flow_matching_loss": 0.050061154365539554, + "grad_norm": 0.40724676847457886, + "learning_rate": 2.9974020642888875e-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": 0.03251873888075352, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.018702817521989344, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.08374029472470283, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.050061154365539554, + "weighted_rkd_loss": 0.041870147362351416 + }, + { + "epoch": 0.587248322147651, + "flow_matching_loss": 0.04783603325486183, + "grad_norm": 0.3322409391403198, + "learning_rate": 2.9967856807370458e-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": 0.02946928795427084, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.016537870932370424, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.07547644712030888, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.04783603325486183, + "weighted_rkd_loss": 0.03773822356015444 + }, + { + "epoch": 0.6040268456375839, + "flow_matching_loss": 0.04893294684588909, + "grad_norm": 0.37779635190963745, + "learning_rate": 2.996103822930615e-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": 0.03286735378205776, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.01721991701051593, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.08295462429523467, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.04893294684588909, + "weighted_rkd_loss": 0.04147731214761734 + }, + { + "epoch": 0.6208053691275168, + "flow_matching_loss": 0.04373093023896217, + "grad_norm": 0.3719472587108612, + "learning_rate": 2.9953565206962653e-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": 0.030151859484612943, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.015161976683884859, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.07546569555997848, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.04373093023896217, + "weighted_rkd_loss": 0.03773284777998924 + }, + { + "epoch": 0.6375838926174496, + "flow_matching_loss": 0.0451244730502367, + "grad_norm": 0.41764843463897705, + "learning_rate": 2.9945438067234172e-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": 0.031219374760985373, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.016777731571346523, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.07921648174524307, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.0451244730502367, + "weighted_rkd_loss": 0.039608240872621533 + }, + { + "epoch": 0.6543624161073825, + "flow_matching_loss": 0.04234543144702911, + "grad_norm": 0.3251534402370453, + "learning_rate": 2.9936657165628124e-05, + "loss": 0.0805, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.03002563174813986, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.01633982127532363, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.07639108411967754, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.04234543144702911, + "weighted_rkd_loss": 0.03819554205983877 + }, + { + "epoch": 0.6711409395973155, + "flow_matching_loss": 0.04180084690451622, + "grad_norm": 0.3378938138484955, + "learning_rate": 2.992722288624959e-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": 0.026612223125994205, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.014572171401232481, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.06779661700129509, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.04180084690451622, + "weighted_rkd_loss": 0.033898308500647546 + }, + { + "epoch": 0.6879194630872483, + "flow_matching_loss": 0.04224371202290058, + "grad_norm": 0.3396628797054291, + "learning_rate": 2.9917135641784515e-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": 0.025039970688521863, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0127363421022892, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.06281628347933292, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.04224371202290058, + "weighted_rkd_loss": 0.03140814173966646 + }, + { + "epoch": 0.7046979865771812, + "flow_matching_loss": 0.03971978519111872, + "grad_norm": 0.30670905113220215, + "learning_rate": 2.9906395873481655e-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": 0.02627517990767956, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.013494232017546893, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.06604459211230278, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.03971978519111872, + "weighted_rkd_loss": 0.03302229605615139 + }, + { + "epoch": 0.7214765100671141, + "flow_matching_loss": 0.04066108614206314, + "grad_norm": 0.39300358295440674, + "learning_rate": 2.989500405113326e-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": 0.026304544880986214, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.01363995224237442, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.06624904237687587, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.04066108614206314, + "weighted_rkd_loss": 0.033124521188437936 + }, + { + "epoch": 0.738255033557047, + "flow_matching_loss": 0.04158542156219482, + "grad_norm": 0.3305646777153015, + "learning_rate": 2.9882960673054547e-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": 0.025151407718658446, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.01327982721850276, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.06358264274895191, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.04158542156219482, + "weighted_rkd_loss": 0.031791321374475955 + }, + { + "epoch": 0.7550335570469798, + "flow_matching_loss": 0.04144757613539696, + "grad_norm": 0.34595179557800293, + "learning_rate": 2.9870266266061886e-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": 0.022307384945452213, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0116576106287539, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.0562723807990551, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.04144757613539696, + "weighted_rkd_loss": 0.02813619039952755 + }, + { + "epoch": 0.7718120805369127, + "flow_matching_loss": 0.04313542582094669, + "grad_norm": 0.3158428370952606, + "learning_rate": 2.9856921385449766e-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": 0.02726253420114517, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.014133189246058464, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.0686582576483488, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.04313542582094669, + "weighted_rkd_loss": 0.0343291288241744 + }, + { + "epoch": 0.7885906040268457, + "flow_matching_loss": 0.042977604642510416, + "grad_norm": 0.3631656765937805, + "learning_rate": 2.9842926614966502e-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": 0.023044148832559584, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.012475840467959642, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.05856413841247558, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.042977604642510416, + "weighted_rkd_loss": 0.02928206920623779 + }, + { + "epoch": 0.8053691275167785, + "flow_matching_loss": 0.03909936882555485, + "grad_norm": 0.29421156644821167, + "learning_rate": 2.982828256678868e-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": 0.02391676288098097, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0120180937461555, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.059851620346307755, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.03909936882555485, + "weighted_rkd_loss": 0.029925810173153877 + }, + { + "epoch": 0.8221476510067114, + "flow_matching_loss": 0.03855758477002382, + "grad_norm": 0.2879815399646759, + "learning_rate": 2.9812989881494412e-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": 0.02446096483618021, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.012417732179164887, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.061339662224054334, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.03855758477002382, + "weighted_rkd_loss": 0.030669831112027167 + }, + { + "epoch": 0.8389261744966443, + "flow_matching_loss": 0.04289716742932796, + "grad_norm": 0.37124210596084595, + "learning_rate": 2.9797049228035296e-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": 0.0244005236774683, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.013081625755876302, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.06188267394900322, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.04289716742932796, + "weighted_rkd_loss": 0.03094133697450161 + }, + { + "epoch": 0.8557046979865772, + "flow_matching_loss": 0.03983708620071411, + "grad_norm": 0.3242025375366211, + "learning_rate": 2.978046130370715e-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": 0.022115133143961428, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.011383708892390132, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.05561397559940815, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.03983708620071411, + "weighted_rkd_loss": 0.027806987799704076 + }, + { + "epoch": 0.87248322147651, + "flow_matching_loss": 0.03861053548753261, + "grad_norm": 0.30338215827941895, + "learning_rate": 2.9763226834119535e-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": 0.021896894834935666, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.011009396938607096, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.054803187027573586, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.03861053548753261, + "weighted_rkd_loss": 0.027401593513786793 + }, + { + "epoch": 0.889261744966443, + "flow_matching_loss": 0.037504769116640094, + "grad_norm": 0.2798364460468292, + "learning_rate": 2.9745346573163963e-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": 0.019946480728685857, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.009748974908143282, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.04964193664491177, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.037504769116640094, + "weighted_rkd_loss": 0.024820968322455884 + }, + { + "epoch": 0.9060402684563759, + "flow_matching_loss": 0.03797667548060417, + "grad_norm": 0.25551819801330566, + "learning_rate": 2.9726821302980987e-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": 0.021121531259268522, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.01155223953537643, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.05379530191421509, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.03797667548060417, + "weighted_rkd_loss": 0.026897650957107545 + }, + { + "epoch": 0.9228187919463087, + "flow_matching_loss": 0.037000850029289725, + "grad_norm": 0.31982243061065674, + "learning_rate": 2.9707651833925935e-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": 0.0204544547945261, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.009314438654109835, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.05022334791719914, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.037000850029289725, + "weighted_rkd_loss": 0.02511167395859957 + }, + { + "epoch": 0.9395973154362416, + "flow_matching_loss": 0.038409977406263354, + "grad_norm": 0.32383039593696594, + "learning_rate": 2.968783900453348e-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": 0.01836938913911581, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.010035208612680434, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.046773987263441084, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.038409977406263354, + "weighted_rkd_loss": 0.023386993631720542 + }, + { + "epoch": 0.9563758389261745, + "flow_matching_loss": 0.03884415850043297, + "grad_norm": 0.3668201267719269, + "learning_rate": 2.9667383681480968e-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": 0.01754778316244483, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.008952949335798622, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.04404851570725441, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.03884415850043297, + "weighted_rkd_loss": 0.022024257853627206 + }, + { + "epoch": 0.9731543624161074, + "flow_matching_loss": 0.03858929816633463, + "grad_norm": 0.31969910860061646, + "learning_rate": 2.9646286759550493e-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": 0.018698493484407663, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.00886254720389843, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.04625953398644924, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.03858929816633463, + "weighted_rkd_loss": 0.02312976699322462 + }, + { + "epoch": 0.9899328859060402, + "flow_matching_loss": 0.037984182313084605, + "grad_norm": 0.37169548869132996, + "learning_rate": 2.9624549161589764e-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": 0.020831290166825056, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.011616774275898934, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.0532793540507555, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.037984182313084605, + "weighted_rkd_loss": 0.02663967702537775 + }, + { + "epoch": 1.0067114093959733, + "flow_matching_loss": 0.032401621155440805, + "grad_norm": 0.2810411751270294, + "learning_rate": 2.9602171838471745e-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": 0.02296459171921015, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.010374387260526419, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.05630357041954994, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.032401621155440805, + "weighted_rkd_loss": 0.02815178520977497 + }, + { + "epoch": 1.023489932885906, + "flow_matching_loss": 0.03286149390041828, + "grad_norm": 0.2521582245826721, + "learning_rate": 2.957915576905303e-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": 0.017456885334104298, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.008559525944292545, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.04347329661250114, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.03286149390041828, + "weighted_rkd_loss": 0.02173664830625057 + }, + { + "epoch": 1.0402684563758389, + "flow_matching_loss": 0.03205829132348299, + "grad_norm": 0.3185127377510071, + "learning_rate": 2.9555501960131058e-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": 0.01915504578500986, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.009150991961359978, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.04746108390390873, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.03205829132348299, + "weighted_rkd_loss": 0.023730541951954363 + }, + { + "epoch": 1.0570469798657718, + "flow_matching_loss": 0.03440665490925312, + "grad_norm": 0.2903192341327667, + "learning_rate": 2.9531211446400068e-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": 0.018206587247550487, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.009964327700436116, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.04637750200927258, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.03440665490925312, + "weighted_rkd_loss": 0.02318875100463629 + }, + { + "epoch": 1.0738255033557047, + "flow_matching_loss": 0.0327774690464139, + "grad_norm": 0.2796550691127777, + "learning_rate": 2.9506285290405813e-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": 0.01788708958774805, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.009063597302883863, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.044837776198983194, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.0327774690464139, + "weighted_rkd_loss": 0.022418888099491597 + }, + { + "epoch": 1.0906040268456376, + "flow_matching_loss": 0.03253862913697958, + "grad_norm": 0.29687777161598206, + "learning_rate": 2.948072458249911e-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": 0.018407645542174576, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.008646867936477065, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.04546215943992138, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.03253862913697958, + "weighted_rkd_loss": 0.02273107971996069 + }, + { + "epoch": 1.1073825503355705, + "flow_matching_loss": 0.034775618463754654, + "grad_norm": 0.26023975014686584, + "learning_rate": 2.9454530440788125e-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": 0.017261851113289595, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.008692631544545294, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.04321633391082287, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.034775618463754654, + "weighted_rkd_loss": 0.021608166955411436 + }, + { + "epoch": 1.1241610738255035, + "flow_matching_loss": 0.03697704430669546, + "grad_norm": 0.2973390817642212, + "learning_rate": 2.9427704011089474e-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": 0.017315912712365388, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.008650409011170267, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.043282233737409116, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.03697704430669546, + "weighted_rkd_loss": 0.021641116868704558 + }, + { + "epoch": 1.1409395973154361, + "flow_matching_loss": 0.034938321262598035, + "grad_norm": 0.33880263566970825, + "learning_rate": 2.94002464668781e-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": 0.01774648055434227, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.008731141500175, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.04422410242259502, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.034938321262598035, + "weighted_rkd_loss": 0.02211205121129751 + }, + { + "epoch": 1.157718120805369, + "flow_matching_loss": 0.03397301379591226, + "grad_norm": 0.30417579412460327, + "learning_rate": 2.9372159009235936e-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": 0.016883174236863852, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0084213403519243, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.04218768924474716, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.03397301379591226, + "weighted_rkd_loss": 0.02109384462237358 + }, + { + "epoch": 1.174496644295302, + "flow_matching_loss": 0.033245801366865634, + "grad_norm": 0.2966938614845276, + "learning_rate": 2.9343442866799365e-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": 0.01730244169011712, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.008166278852149845, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.04277116190642118, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.033245801366865634, + "weighted_rkd_loss": 0.02138558095321059 + }, + { + "epoch": 1.191275167785235, + "flow_matching_loss": 0.032241361774504186, + "grad_norm": 0.3242097496986389, + "learning_rate": 2.9314099295705495e-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": 0.01751971701160073, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.008714311430230737, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.04375374540686607, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.032241361774504186, + "weighted_rkd_loss": 0.021876872703433036 + }, + { + "epoch": 1.2080536912751678, + "flow_matching_loss": 0.030840910412371157, + "grad_norm": 0.3416939675807953, + "learning_rate": 2.928412957953718e-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": 0.016577553190290927, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.008127822168171405, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.041282928362488744, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.030840910412371157, + "weighted_rkd_loss": 0.020641464181244372 + }, + { + "epoch": 1.2248322147651007, + "flow_matching_loss": 0.030589586496353148, + "grad_norm": 0.3434118926525116, + "learning_rate": 2.925353502926689e-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": 0.014619570132344962, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.00703063509427011, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.03626977521926165, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.030589586496353148, + "weighted_rkd_loss": 0.018134887609630824 + }, + { + "epoch": 1.2416107382550337, + "flow_matching_loss": 0.028109255991876126, + "grad_norm": 0.2770143449306488, + "learning_rate": 2.922231698319937e-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": 0.015383913926780225, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.007206043181940913, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.03797387108206749, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.028109255991876126, + "weighted_rkd_loss": 0.018986935541033746 + }, + { + "epoch": 1.2583892617449663, + "flow_matching_loss": 0.03261192515492439, + "grad_norm": 0.271056592464447, + "learning_rate": 2.9190476806913088e-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": 0.016976217925548553, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.007505776965990663, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.04145821258425712, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.03261192515492439, + "weighted_rkd_loss": 0.02072910629212856 + }, + { + "epoch": 1.2751677852348993, + "flow_matching_loss": 0.030070590786635876, + "grad_norm": 0.2317705899477005, + "learning_rate": 2.9158015893200504e-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": 0.016096030082553624, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.007664719317108393, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.03985677920281887, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.030070590786635876, + "weighted_rkd_loss": 0.019928389601409434 + }, + { + "epoch": 1.2919463087248322, + "flow_matching_loss": 0.033119112625718115, + "grad_norm": 0.2506316006183624, + "learning_rate": 2.9124935662007137e-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": 0.01613023495301604, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.007737922063097358, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.03999839182943106, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.033119112625718115, + "weighted_rkd_loss": 0.01999919591471553 + }, + { + "epoch": 1.308724832214765, + "flow_matching_loss": 0.031553388386964795, + "grad_norm": 0.2889447808265686, + "learning_rate": 2.9091237560369467e-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": 0.01787243187427521, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.009260696731507779, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.04500555992126465, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.031553388386964795, + "weighted_rkd_loss": 0.022502779960632324 + }, + { + "epoch": 1.325503355704698, + "flow_matching_loss": 0.03222518637776375, + "grad_norm": 0.2785235047340393, + "learning_rate": 2.905692306235163e-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": 0.015721357706934214, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.007582925446331501, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.039025640860199926, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.03222518637776375, + "weighted_rkd_loss": 0.019512820430099963 + }, + { + "epoch": 1.342281879194631, + "flow_matching_loss": 0.030334611982107164, + "grad_norm": 0.31459397077560425, + "learning_rate": 2.9021993668980925e-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": 0.01560999769717455, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.007159262569621206, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.03837925773113966, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.030334611982107164, + "weighted_rkd_loss": 0.01918962886556983 + }, + { + "epoch": 1.3590604026845639, + "flow_matching_loss": 0.03257175628095865, + "grad_norm": 0.2993733286857605, + "learning_rate": 2.8986450908182185e-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": 0.015799124166369437, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.00794396479614079, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.039542213454842566, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.03257175628095865, + "weighted_rkd_loss": 0.019771106727421283 + }, + { + "epoch": 1.3758389261744965, + "flow_matching_loss": 0.032540197111666204, + "grad_norm": 0.31480374932289124, + "learning_rate": 2.8950296334710902e-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": 0.01558954007923603, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.007458786852657795, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.03863786738365889, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.032540197111666204, + "weighted_rkd_loss": 0.019318933691829444 + }, + { + "epoch": 1.3926174496644295, + "flow_matching_loss": 0.03170314058661461, + "grad_norm": 0.2934443950653076, + "learning_rate": 2.8913531530085248e-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": 0.016132532991468905, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.007935636956244708, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.040200702473521234, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.03170314058661461, + "weighted_rkd_loss": 0.020100351236760617 + }, + { + "epoch": 1.4093959731543624, + "flow_matching_loss": 0.03200640454888344, + "grad_norm": 0.26509004831314087, + "learning_rate": 2.887615810251687e-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": 0.014211549051105976, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.006883570412173867, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.035306668281555174, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.03200640454888344, + "weighted_rkd_loss": 0.017653334140777587 + }, + { + "epoch": 1.4261744966442953, + "flow_matching_loss": 0.02827311120927334, + "grad_norm": 0.3237679600715637, + "learning_rate": 2.8838177686840577e-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": 0.017118581011891364, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.00820444175042212, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.04244160372763872, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.02827311120927334, + "weighted_rkd_loss": 0.02122080186381936 + }, + { + "epoch": 1.4429530201342282, + "flow_matching_loss": 0.03121668491512537, + "grad_norm": 0.3042275607585907, + "learning_rate": 2.8799591944442776e-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": 0.016424584202468395, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.007744280295446515, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.040593449398875235, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.03121668491512537, + "weighted_rkd_loss": 0.020296724699437618 + }, + { + "epoch": 1.4597315436241611, + "flow_matching_loss": 0.030370122380554677, + "grad_norm": 0.3472956120967865, + "learning_rate": 2.8760402563188848e-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": 0.017987641133368017, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.009500452037900686, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.045475734397768974, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.030370122380554677, + "weighted_rkd_loss": 0.022737867198884487 + }, + { + "epoch": 1.476510067114094, + "flow_matching_loss": 0.02861998286098242, + "grad_norm": 0.24505336582660675, + "learning_rate": 2.8720611257349285e-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": 0.0141552341170609, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.006439156364649534, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.03474962469190359, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.02861998286098242, + "weighted_rkd_loss": 0.017374812345951795 + }, + { + "epoch": 1.4932885906040267, + "flow_matching_loss": 0.03184323962777853, + "grad_norm": 0.2608340084552765, + "learning_rate": 2.868021976752471e-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": 0.01494277287274599, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.00734082362614572, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.037226369604468346, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.03184323962777853, + "weighted_rkd_loss": 0.018613184802234173 + }, + { + "epoch": 1.5100671140939599, + "flow_matching_loss": 0.03042609766125679, + "grad_norm": 0.333564430475235, + "learning_rate": 2.8639229860569738e-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": 0.0160973247140646, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.007541899429634213, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.03973654881119728, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.03042609766125679, + "weighted_rkd_loss": 0.01986827440559864 + }, + { + "epoch": 1.5268456375838926, + "flow_matching_loss": 0.03012043833732605, + "grad_norm": 0.24364671111106873, + "learning_rate": 2.8597643329515688e-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": 0.014172045607119798, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.006650603329762817, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.034994694776833056, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.03012043833732605, + "weighted_rkd_loss": 0.017497347388416528 + }, + { + "epoch": 1.5436241610738255, + "flow_matching_loss": 0.03092307262122631, + "grad_norm": 0.27232593297958374, + "learning_rate": 2.855546199349214e-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": 0.014154891762882472, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.007071852730587125, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.03538163620978594, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.03092307262122631, + "weighted_rkd_loss": 0.01769081810489297 + }, + { + "epoch": 1.5604026845637584, + "flow_matching_loss": 0.027929746173322202, + "grad_norm": 0.33788374066352844, + "learning_rate": 2.85126876976474e-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": 0.012748553790152073, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.005962743982672691, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.031459851376712325, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.027929746173322202, + "weighted_rkd_loss": 0.015729925688356163 + }, + { + "epoch": 1.5771812080536913, + "flow_matching_loss": 0.028270901925861836, + "grad_norm": 0.21897122263908386, + "learning_rate": 2.8469322313067726e-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": 0.015070199593901634, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.007116558589041233, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.0372569577768445, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.028270901925861836, + "weighted_rkd_loss": 0.01862847888842225 + }, + { + "epoch": 1.5939597315436242, + "flow_matching_loss": 0.02921714950352907, + "grad_norm": 0.3024495542049408, + "learning_rate": 2.8425367736695522e-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": 0.011751286219805479, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.005542754894122481, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.02904532738029957, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.02921714950352907, + "weighted_rkd_loss": 0.014522663690149784 + }, + { + "epoch": 1.610738255033557, + "flow_matching_loss": 0.02695902418345213, + "grad_norm": 0.2782779335975647, + "learning_rate": 2.838082589124636e-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": 0.015185461472719907, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.007375222817063332, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.03774614538997412, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.02695902418345213, + "weighted_rkd_loss": 0.01887307269498706 + }, + { + "epoch": 1.62751677852349, + "flow_matching_loss": 0.026085390895605087, + "grad_norm": 0.31670045852661133, + "learning_rate": 2.8335698725124843e-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": 0.014663305971771479, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.007240531407296658, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.03656714335083962, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.026085390895605087, + "weighted_rkd_loss": 0.01828357167541981 + }, + { + "epoch": 1.6442953020134228, + "flow_matching_loss": 0.02555117029696703, + "grad_norm": 0.2121639996767044, + "learning_rate": 2.8289988212339422e-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": 0.013465657457709312, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0062181116081774235, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.03314942680299282, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.02555117029696703, + "weighted_rkd_loss": 0.01657471340149641 + }, + { + "epoch": 1.6610738255033557, + "flow_matching_loss": 0.026020037569105627, + "grad_norm": 0.22167570888996124, + "learning_rate": 2.8243696352415997e-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": 0.01538208145648241, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.008201855747029186, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.038966018334031106, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.026020037569105627, + "weighted_rkd_loss": 0.019483009167015553 + }, + { + "epoch": 1.6778523489932886, + "flow_matching_loss": 0.029859650880098343, + "grad_norm": 0.2646207809448242, + "learning_rate": 2.8196825170310495e-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": 0.012622483912855387, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.005989065254107117, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.031234032846987246, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.029859650880098343, + "weighted_rkd_loss": 0.015617016423493623 + }, + { + "epoch": 1.6946308724832215, + "flow_matching_loss": 0.029160509817302227, + "grad_norm": 0.29436981678009033, + "learning_rate": 2.8149376716320256e-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": 0.014717372134327888, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.006617354042828083, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.036052098497748375, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.029160509817302227, + "weighted_rkd_loss": 0.018026049248874187 + }, + { + "epoch": 1.7114093959731544, + "flow_matching_loss": 0.02514624744653702, + "grad_norm": 0.2813986837863922, + "learning_rate": 2.8101353065994373e-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": 0.014104087930172681, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.007160045579075814, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.03536822162568569, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.02514624744653702, + "weighted_rkd_loss": 0.017684110812842845 + }, + { + "epoch": 1.7281879194630871, + "flow_matching_loss": 0.025560924783349037, + "grad_norm": 0.24623122811317444, + "learning_rate": 2.8052756320042885e-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": 0.013708307594060897, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.006159724202007055, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.03357633929699659, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.025560924783349037, + "weighted_rkd_loss": 0.016788169648498295 + }, + { + "epoch": 1.7449664429530203, + "flow_matching_loss": 0.02597265727818012, + "grad_norm": 0.24410276114940643, + "learning_rate": 2.8003588604244893e-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": 0.014219438191503286, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.006340369349345565, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.0347792461514473, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.02597265727818012, + "weighted_rkd_loss": 0.01738962307572365 + }, + { + "epoch": 1.761744966442953, + "flow_matching_loss": 0.02521944120526314, + "grad_norm": 0.24740788340568542, + "learning_rate": 2.7953852069355565e-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": 0.014595606923103332, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0072414461523294445, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.036432659812271596, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.02521944120526314, + "weighted_rkd_loss": 0.018216329906135798 + }, + { + "epoch": 1.778523489932886, + "flow_matching_loss": 0.027643476985394953, + "grad_norm": 0.3433672785758972, + "learning_rate": 2.790354889101206e-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": 0.013499390333890915, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.006195838330313563, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.03319461885839701, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.027643476985394953, + "weighted_rkd_loss": 0.016597309429198504 + }, + { + "epoch": 1.7953020134228188, + "flow_matching_loss": 0.02857300378382206, + "grad_norm": 0.24258039891719818, + "learning_rate": 2.7852681269638366e-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": 0.013103808369487524, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.006252429843880236, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.03246004665270448, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.02857300378382206, + "weighted_rkd_loss": 0.01623002332635224 + }, + { + "epoch": 1.8120805369127517, + "flow_matching_loss": 0.025308608263731002, + "grad_norm": 0.25732696056365967, + "learning_rate": 2.780125143034902e-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": 0.014380837325006723, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.006497780932113529, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.03525945525616407, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.025308608263731002, + "weighted_rkd_loss": 0.017629727628082036 + }, + { + "epoch": 1.8288590604026846, + "flow_matching_loss": 0.026384789682924747, + "grad_norm": 0.32120704650878906, + "learning_rate": 2.7749261622851812e-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": 0.01262503294274211, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.006175780994817614, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.031425846926867965, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.026384789682924747, + "weighted_rkd_loss": 0.015712923463433982 + }, + { + "epoch": 1.8456375838926173, + "flow_matching_loss": 0.027396404929459096, + "grad_norm": 0.21401037275791168, + "learning_rate": 2.7696714121349335e-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": 0.011610002536326647, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.005550801800563931, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.028770807012915612, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.027396404929459096, + "weighted_rkd_loss": 0.014385403506457806 + }, + { + "epoch": 1.8624161073825505, + "flow_matching_loss": 0.0279746038839221, + "grad_norm": 0.20054933428764343, + "learning_rate": 2.764361122443954e-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": 0.011772450804710389, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.005359048675745726, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.028903950192034244, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.0279746038839221, + "weighted_rkd_loss": 0.014451975096017122 + }, + { + "epoch": 1.8791946308724832, + "flow_matching_loss": 0.02448360715061426, + "grad_norm": 0.26301416754722595, + "learning_rate": 2.758995525501517e-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": 0.011649888660758733, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.00557361776009202, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.02887339498847723, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.02448360715061426, + "weighted_rkd_loss": 0.014436697494238615 + }, + { + "epoch": 1.895973154362416, + "flow_matching_loss": 0.027417019382119177, + "grad_norm": 0.22313368320465088, + "learning_rate": 2.753574856016215e-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": 0.012046439200639724, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.005405143462121486, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.029498021490871905, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.027417019382119177, + "weighted_rkd_loss": 0.014749010745435952 + }, + { + "epoch": 1.912751677852349, + "flow_matching_loss": 0.02374915238469839, + "grad_norm": 0.2883581817150116, + "learning_rate": 2.7480993511056913e-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": 0.013785356283187866, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0063822722062468525, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.03395298458635807, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.02374915238469839, + "weighted_rkd_loss": 0.016976492293179036 + }, + { + "epoch": 1.929530201342282, + "flow_matching_loss": 0.024885526299476622, + "grad_norm": 0.2205721139907837, + "learning_rate": 2.7425692502862684e-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": 0.012991191726177931, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.006324190460145473, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.032306574285030365, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.024885526299476622, + "weighted_rkd_loss": 0.016153287142515182 + }, + { + "epoch": 1.9463087248322148, + "flow_matching_loss": 0.027140559628605843, + "grad_norm": 0.32631516456604004, + "learning_rate": 2.7369847954624728e-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": 0.01370504880324006, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.006441503390669822, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.03385160099714994, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.027140559628605843, + "weighted_rkd_loss": 0.01692580049857497 + }, + { + "epoch": 1.9630872483221475, + "flow_matching_loss": 0.027510961145162584, + "grad_norm": 0.19017235934734344, + "learning_rate": 2.731346230916448e-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": 0.010960731562227011, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0051203057635575535, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.02704176902770996, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.027510961145162584, + "weighted_rkd_loss": 0.01352088451385498 + }, + { + "epoch": 1.9798657718120807, + "flow_matching_loss": 0.023988731019198895, + "grad_norm": 0.2913515865802765, + "learning_rate": 2.7256538032972746e-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": 0.010911648208275438, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0053501016926020386, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.0271733982488513, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.023988731019198895, + "weighted_rkd_loss": 0.01358669912442565 + }, + { + "epoch": 1.9966442953020134, + "flow_matching_loss": 0.023296417482197285, + "grad_norm": 0.2592824697494507, + "learning_rate": 2.7199077616101773e-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": 0.01110992943868041, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.004422898893244565, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.026642757840454577, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.023296417482197285, + "weighted_rkd_loss": 0.013321378920227289 + }, + { + "epoch": 2.0134228187919465, + "flow_matching_loss": 0.024605618230998515, + "grad_norm": 0.3102531135082245, + "learning_rate": 2.714108357205634e-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": 0.01258283476345241, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0052482510916888716, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.03041392080485821, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.024605618230998515, + "weighted_rkd_loss": 0.015206960402429104 + }, + { + "epoch": 2.030201342281879, + "flow_matching_loss": 0.027245542779564858, + "grad_norm": 0.22417199611663818, + "learning_rate": 2.7082558437683803e-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": 0.01143276859074831, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.005770996958017349, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.028636534698307514, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.027245542779564858, + "weighted_rkd_loss": 0.014318267349153757 + }, + { + "epoch": 2.046979865771812, + "flow_matching_loss": 0.023954916559159756, + "grad_norm": 0.20610450208187103, + "learning_rate": 2.702350477306315e-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": 0.011581839248538017, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.005343518941663206, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.028507197834551336, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.023954916559159756, + "weighted_rkd_loss": 0.014253598917275668 + }, + { + "epoch": 2.063758389261745, + "flow_matching_loss": 0.023449451476335526, + "grad_norm": 0.2560037672519684, + "learning_rate": 2.6963925161392963e-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": 0.010292424075305461, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.004632242536172271, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.02521709054708481, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.023449451476335526, + "weighted_rkd_loss": 0.012608545273542405 + }, + { + "epoch": 2.0805369127516777, + "flow_matching_loss": 0.02585367038846016, + "grad_norm": 0.23351788520812988, + "learning_rate": 2.6903822208878476e-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": 0.013421973027288914, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.005883032223209739, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.032726977951824665, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.02585367038846016, + "weighted_rkd_loss": 0.016363488975912333 + }, + { + "epoch": 2.097315436241611, + "flow_matching_loss": 0.02222144976258278, + "grad_norm": 0.19517207145690918, + "learning_rate": 2.6843198544617535e-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": 0.010299140820279717, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0048344668466597795, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.025432748906314372, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.02222144976258278, + "weighted_rkd_loss": 0.012716374453157186 + }, + { + "epoch": 2.1140939597315436, + "flow_matching_loss": 0.026821971125900745, + "grad_norm": 0.3056233823299408, + "learning_rate": 2.6782056820485612e-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": 0.013456127140671015, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.006366095319390297, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.03327834941446781, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.026821971125900745, + "weighted_rkd_loss": 0.016639174707233906 + }, + { + "epoch": 2.1308724832214767, + "flow_matching_loss": 0.024398398119956255, + "grad_norm": 0.2585718035697937, + "learning_rate": 2.672039971101979e-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": 0.011472947150468826, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.004935935512185097, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.027881830371916293, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.024398398119956255, + "weighted_rkd_loss": 0.013940915185958146 + }, + { + "epoch": 2.1476510067114094, + "flow_matching_loss": 0.024629255197942256, + "grad_norm": 0.1899724155664444, + "learning_rate": 2.665822991330178e-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": 0.010962147265672684, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.004843188123777509, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.026767482981085777, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.024629255197942256, + "weighted_rkd_loss": 0.013383741490542889 + }, + { + "epoch": 2.164429530201342, + "flow_matching_loss": 0.02464695330709219, + "grad_norm": 0.19738611578941345, + "learning_rate": 2.6595550146839934e-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": 0.011003640946000814, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0048983822576701645, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.026905664429068565, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.02464695330709219, + "weighted_rkd_loss": 0.013452832214534282 + }, + { + "epoch": 2.1812080536912752, + "flow_matching_loss": 0.024324053525924684, + "grad_norm": 0.21592102944850922, + "learning_rate": 2.6532363153450287e-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": 0.011682957503944635, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.00482977582141757, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.0281956909224391, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.024324053525924684, + "weighted_rkd_loss": 0.01409784546121955 + }, + { + "epoch": 2.197986577181208, + "flow_matching_loss": 0.02110758163034916, + "grad_norm": 0.18565689027309418, + "learning_rate": 2.6468671697136634e-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": 0.010184981161728501, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.004276543972082436, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.024646506272256376, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.02110758163034916, + "weighted_rkd_loss": 0.012323253136128188 + }, + { + "epoch": 2.214765100671141, + "flow_matching_loss": 0.024282259680330754, + "grad_norm": 0.2649759352207184, + "learning_rate": 2.640447856396961e-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": 0.012323899054899811, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.005656029563397169, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.030303827580064537, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.024282259680330754, + "weighted_rkd_loss": 0.015151913790032268 + }, + { + "epoch": 2.2315436241610738, + "flow_matching_loss": 0.02368467003107071, + "grad_norm": 0.1718354970216751, + "learning_rate": 2.6339786561964805e-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": 0.010672940826043486, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.005090259737335146, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.02643614150583744, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.02368467003107071, + "weighted_rkd_loss": 0.01321807075291872 + }, + { + "epoch": 2.248322147651007, + "flow_matching_loss": 0.02225311063230038, + "grad_norm": 0.1958433985710144, + "learning_rate": 2.6274598520959966e-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": 0.010571935120970011, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.004792783199809492, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.02593665346503258, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.02225311063230038, + "weighted_rkd_loss": 0.01296832673251629 + }, + { + "epoch": 2.2651006711409396, + "flow_matching_loss": 0.022798995673656463, + "grad_norm": 0.2145613580942154, + "learning_rate": 2.6208917292491185e-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": 0.010735450917854905, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.005248784064315259, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.026719686202704908, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.022798995673656463, + "weighted_rkd_loss": 0.013359843101352454 + }, + { + "epoch": 2.2818791946308723, + "flow_matching_loss": 0.0219331169500947, + "grad_norm": 0.18633633852005005, + "learning_rate": 2.6142745749668176e-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": 0.011048201052471996, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.005030419118702412, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.02712682094424963, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.0219331169500947, + "weighted_rkd_loss": 0.013563410472124815 + }, + { + "epoch": 2.2986577181208054, + "flow_matching_loss": 0.021648623794317246, + "grad_norm": 0.33190497756004333, + "learning_rate": 2.607608678704859e-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": 0.01214893334545195, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.005797731433995068, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.030095598101615904, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.021648623794317246, + "weighted_rkd_loss": 0.015047799050807952 + }, + { + "epoch": 2.315436241610738, + "flow_matching_loss": 0.022526911459863186, + "grad_norm": 0.23687943816184998, + "learning_rate": 2.600894332051139e-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": 0.009939734917134046, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.004744947282597422, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.024624417070299387, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.022526911459863186, + "weighted_rkd_loss": 0.012312208535149693 + }, + { + "epoch": 2.3322147651006713, + "flow_matching_loss": 0.02625675145536661, + "grad_norm": 0.2014767974615097, + "learning_rate": 2.594131828712932e-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": 0.011044638883322477, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0052457981742918495, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.02733507603406906, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.02625675145536661, + "weighted_rkd_loss": 0.01366753801703453 + }, + { + "epoch": 2.348993288590604, + "flow_matching_loss": 0.023217788338661192, + "grad_norm": 0.30442988872528076, + "learning_rate": 2.5873214645040414e-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": 0.01133463829755783, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.005828363331966102, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.028497640043497086, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.023217788338661192, + "weighted_rkd_loss": 0.014248820021748543 + }, + { + "epoch": 2.365771812080537, + "flow_matching_loss": 0.023031603731215, + "grad_norm": 0.25246021151542664, + "learning_rate": 2.5804635373318604e-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": 0.009690780472010375, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.004492142889648676, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.023873704113066198, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.023031603731215, + "weighted_rkd_loss": 0.011936852056533099 + }, + { + "epoch": 2.38255033557047, + "flow_matching_loss": 0.024585946835577487, + "grad_norm": 0.20293375849723816, + "learning_rate": 2.57355834718434e-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": 0.011088294768705964, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0044671180192381145, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.02664370760321617, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.024585946835577487, + "weighted_rkd_loss": 0.013321853801608085 + }, + { + "epoch": 2.3993288590604025, + "flow_matching_loss": 0.021987343579530715, + "grad_norm": 0.23088401556015015, + "learning_rate": 2.566606196116867e-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": 0.010794766712933779, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.004875816567800939, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.026465350203216075, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.021987343579530715, + "weighted_rkd_loss": 0.013232675101608037 + }, + { + "epoch": 2.4161073825503356, + "flow_matching_loss": 0.024937476590275766, + "grad_norm": 0.21601395308971405, + "learning_rate": 2.559607388239051e-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": 0.01238170089200139, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.005331674730405212, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.03009507693350315, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.024937476590275766, + "weighted_rkd_loss": 0.015047538466751576 + }, + { + "epoch": 2.4328859060402683, + "flow_matching_loss": 0.020688330102711917, + "grad_norm": 0.22601929306983948, + "learning_rate": 2.552562229701422e-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": 0.010842163767665624, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0046914838254451755, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.02637581154704094, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.020688330102711917, + "weighted_rkd_loss": 0.01318790577352047 + }, + { + "epoch": 2.4496644295302015, + "flow_matching_loss": 0.02394238654524088, + "grad_norm": 0.2014225870370865, + "learning_rate": 2.545471028682036e-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": 0.009968520933762192, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.004925135779194534, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.024862177856266498, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.02394238654524088, + "weighted_rkd_loss": 0.012431088928133249 + }, + { + "epoch": 2.466442953020134, + "flow_matching_loss": 0.02508413475006819, + "grad_norm": 0.28125983476638794, + "learning_rate": 2.538334095373e-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": 0.011306649027392267, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.005316834663972259, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.027930132672190668, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.02508413475006819, + "weighted_rkd_loss": 0.013965066336095334 + }, + { + "epoch": 2.4832214765100673, + "flow_matching_loss": 0.023891801387071608, + "grad_norm": 0.29100310802459717, + "learning_rate": 2.531151741966896e-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": 0.01073241732083261, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.005184279289096594, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.026649114117026328, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.023891801387071608, + "weighted_rkd_loss": 0.013324557058513164 + }, + { + "epoch": 2.5, + "flow_matching_loss": 0.02415292225778103, + "grad_norm": 0.22286556661128998, + "learning_rate": 2.5239242826431302e-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": 0.010552192945033312, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.004753443133085966, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.025857829488813877, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.02415292225778103, + "weighted_rkd_loss": 0.012928914744406939 + }, + { + "epoch": 2.5167785234899327, + "flow_matching_loss": 0.022898601554334163, + "grad_norm": 0.18709388375282288, + "learning_rate": 2.5166520335541865e-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": 0.01060750326141715, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.004809713829308749, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.02602472063153982, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.022898601554334163, + "weighted_rkd_loss": 0.01301236031576991 + }, + { + "epoch": 2.533557046979866, + "flow_matching_loss": 0.02009684108197689, + "grad_norm": 0.25758233666419983, + "learning_rate": 2.5093353128117996e-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": 0.010353196039795875, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.00480845645070076, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.025514848809689284, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.02009684108197689, + "weighted_rkd_loss": 0.012757424404844642 + }, + { + "epoch": 2.5503355704697985, + "flow_matching_loss": 0.02130047921091318, + "grad_norm": 0.16162274777889252, + "learning_rate": 2.501974440473037e-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": 0.009250646131113172, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0046441263752058145, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.023145418614149094, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.02130047921091318, + "weighted_rkd_loss": 0.011572709307074547 + }, + { + "epoch": 2.5671140939597317, + "flow_matching_loss": 0.0252719109877944, + "grad_norm": 0.21416917443275452, + "learning_rate": 2.4945697385263015e-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": 0.009564307797700167, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.004446143656969071, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.023574758879840375, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.0252719109877944, + "weighted_rkd_loss": 0.011787379439920188 + }, + { + "epoch": 2.5838926174496644, + "flow_matching_loss": 0.022078343108296393, + "grad_norm": 0.21463052928447723, + "learning_rate": 2.487121530877243e-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": 0.011745102750137449, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.005604003882035613, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.02909420933574438, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.022078343108296393, + "weighted_rkd_loss": 0.01454710466787219 + }, + { + "epoch": 2.600671140939597, + "flow_matching_loss": 0.021502092573791744, + "grad_norm": 0.23142777383327484, + "learning_rate": 2.4796301433345943e-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": 0.0092654038220644, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0041564338142052295, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.022687241435050964, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.021502092573791744, + "weighted_rkd_loss": 0.011343620717525482 + }, + { + "epoch": 2.61744966442953, + "flow_matching_loss": 0.022588416188955306, + "grad_norm": 0.2171335220336914, + "learning_rate": 2.4720959035959148e-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": 0.010664369678124785, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.004696646006777883, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.02602538522332907, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.022588416188955306, + "weighted_rkd_loss": 0.013012692611664534 + }, + { + "epoch": 2.6342281879194633, + "flow_matching_loss": 0.022199026308953762, + "grad_norm": 0.2863241732120514, + "learning_rate": 2.464519141233257e-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": 0.011123205535113811, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0050924249924719335, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.027338836155831815, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.022199026308953762, + "weighted_rkd_loss": 0.013669418077915907 + }, + { + "epoch": 2.651006711409396, + "flow_matching_loss": 0.02178021539002657, + "grad_norm": 0.23810994625091553, + "learning_rate": 2.456900187678754e-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": 0.010764320380985737, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.005336681194603443, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.026865322329103946, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.02178021539002657, + "weighted_rkd_loss": 0.013432661164551973 + }, + { + "epoch": 2.6677852348993287, + "flow_matching_loss": 0.022027206886559726, + "grad_norm": 0.17385552823543549, + "learning_rate": 2.4492393762101147e-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": 0.008514114422723651, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.004043441731482744, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.021071670670062305, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.022027206886559726, + "weighted_rkd_loss": 0.010535835335031152 + }, + { + "epoch": 2.684563758389262, + "flow_matching_loss": 0.022792240604758263, + "grad_norm": 0.20645810663700104, + "learning_rate": 2.4415370419360508e-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": 0.00922980452887714, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.004303267924115062, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.02276287693530321, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.022792240604758263, + "weighted_rkd_loss": 0.011381438467651606 + }, + { + "epoch": 2.7013422818791946, + "flow_matching_loss": 0.021514329593628645, + "grad_norm": 0.2190205603837967, + "learning_rate": 2.4337935217816157e-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": 0.009895919263362885, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.00468641291372478, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.02447825139388442, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.021514329593628645, + "weighted_rkd_loss": 0.01223912569694221 + }, + { + "epoch": 2.7181208053691277, + "flow_matching_loss": 0.024532474111765623, + "grad_norm": 0.3014610707759857, + "learning_rate": 2.4260091544734665e-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": 0.010453154426068067, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0046753128292039035, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.02558162175118923, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.024532474111765623, + "weighted_rkd_loss": 0.012790810875594616 + }, + { + "epoch": 2.7348993288590604, + "flow_matching_loss": 0.022621615324169397, + "grad_norm": 0.20125490427017212, + "learning_rate": 2.4181842805250466e-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": 0.008896962832659483, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.004234325815923512, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.022028251271694898, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.022621615324169397, + "weighted_rkd_loss": 0.011014125635847449 + }, + { + "epoch": 2.751677852348993, + "flow_matching_loss": 0.021766430512070657, + "grad_norm": 0.21964611113071442, + "learning_rate": 2.4103192422216923e-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": 0.009006364457309245, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.003936750791035592, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.021949479822069407, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.021766430512070657, + "weighted_rkd_loss": 0.010974739911034703 + }, + { + "epoch": 2.7684563758389262, + "flow_matching_loss": 0.022125181183218955, + "grad_norm": 0.18881672620773315, + "learning_rate": 2.402414383605658e-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": 0.009801744995638729, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.004149720002897084, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.023753209877759217, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.022125181183218955, + "weighted_rkd_loss": 0.011876604938879608 + }, + { + "epoch": 2.785234899328859, + "flow_matching_loss": 0.022959327884018423, + "grad_norm": 0.20707227289676666, + "learning_rate": 2.3944700504610674e-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": 0.010146900825202466, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.004097336274571717, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.024391137808561326, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.022959327884018423, + "weighted_rkd_loss": 0.012195568904280663 + }, + { + "epoch": 2.802013422818792, + "flow_matching_loss": 0.02108875922858715, + "grad_norm": 0.20151439309120178, + "learning_rate": 2.3864865902987883e-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": 0.008188423234969378, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.003603689349256456, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.019980535749346016, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.02108875922858715, + "weighted_rkd_loss": 0.009990267874673008 + }, + { + "epoch": 2.8187919463087248, + "flow_matching_loss": 0.02431239653378725, + "grad_norm": 0.17307604849338531, + "learning_rate": 2.378464352341231e-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": 0.00882768682204187, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.004448901535943151, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.02210427522659302, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.02431239653378725, + "weighted_rkd_loss": 0.01105213761329651 + }, + { + "epoch": 2.8355704697986575, + "flow_matching_loss": 0.021007327269762753, + "grad_norm": 0.1367725431919098, + "learning_rate": 2.3704036875070728e-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": 0.008387606404721738, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0035082216607406737, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.020283434353768826, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.021007327269762753, + "weighted_rkd_loss": 0.010141717176884413 + }, + { + "epoch": 2.8523489932885906, + "flow_matching_loss": 0.022363264858722687, + "grad_norm": 0.21401777863502502, + "learning_rate": 2.362304948395906e-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": 0.008863104181364179, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.004238870576955378, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.0219650786370039, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.022363264858722687, + "weighted_rkd_loss": 0.01098253931850195 + }, + { + "epoch": 2.8691275167785237, + "flow_matching_loss": 0.021167028322815896, + "grad_norm": 0.2148161679506302, + "learning_rate": 2.3541684892728157e-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": 0.008788073668256402, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.004058226011693478, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.021634373534470796, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.021167028322815896, + "weighted_rkd_loss": 0.010817186767235398 + }, + { + "epoch": 2.8859060402684564, + "flow_matching_loss": 0.02436531288549304, + "grad_norm": 0.21661841869354248, + "learning_rate": 2.345994666052883e-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": 0.010374712059274316, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.004931548284366727, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.025680972170084714, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.02436531288549304, + "weighted_rkd_loss": 0.012840486085042357 + }, + { + "epoch": 2.902684563758389, + "flow_matching_loss": 0.022530085407197475, + "grad_norm": 0.17632728815078735, + "learning_rate": 2.337783836285614e-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": 0.008637029910460114, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0039325655903667215, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.021206625550985337, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.022530085407197475, + "weighted_rkd_loss": 0.010603312775492668 + }, + { + "epoch": 2.9194630872483223, + "flow_matching_loss": 0.020638199150562288, + "grad_norm": 0.21533295512199402, + "learning_rate": 2.3295363591393026e-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": 0.009914667624980212, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.004476642981171608, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.024305978044867515, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.020638199150562288, + "weighted_rkd_loss": 0.012152989022433757 + }, + { + "epoch": 2.936241610738255, + "flow_matching_loss": 0.022945385053753854, + "grad_norm": 0.24226795136928558, + "learning_rate": 2.3212525953853174e-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": 0.009643034590408207, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.005037153465673327, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.024323222693055868, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.022945385053753854, + "weighted_rkd_loss": 0.012161611346527934 + }, + { + "epoch": 2.953020134228188, + "flow_matching_loss": 0.021384773962199687, + "grad_norm": 0.20606893301010132, + "learning_rate": 2.3129329073823206e-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": 0.008418774185702204, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0036228498327545823, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.020460398215800524, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.021384773962199687, + "weighted_rkd_loss": 0.010230199107900262 + }, + { + "epoch": 2.969798657718121, + "flow_matching_loss": 0.02141566090285778, + "grad_norm": 0.21554461121559143, + "learning_rate": 2.3045776590604175e-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": 0.010217978153377771, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.004908383940346539, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.025344339944422246, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.02141566090285778, + "weighted_rkd_loss": 0.012672169972211123 + }, + { + "epoch": 2.9865771812080535, + "flow_matching_loss": 0.02188427047803998, + "grad_norm": 0.23835872113704681, + "learning_rate": 2.296187215905237e-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": 0.008762435987591743, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.00426900724414736, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.021793879009783267, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.02188427047803998, + "weighted_rkd_loss": 0.010896939504891633 + }, + { + "epoch": 3.0033557046979866, + "flow_matching_loss": 0.022038573306053875, + "grad_norm": 0.1795431524515152, + "learning_rate": 2.2877619449419437e-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": 0.007520042033866048, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.003554393211379647, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.018594477139413357, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.022038573306053875, + "weighted_rkd_loss": 0.009297238569706679 + }, + { + "epoch": 3.0201342281879193, + "flow_matching_loss": 0.02141010984778404, + "grad_norm": 0.2697470486164093, + "learning_rate": 2.2793022147191843e-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": 0.009075135085731744, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.004234813805669546, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.022385083977133034, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.02141010984778404, + "weighted_rkd_loss": 0.011192541988566517 + }, + { + "epoch": 3.0369127516778525, + "flow_matching_loss": 0.02096653329208493, + "grad_norm": 0.18201664090156555, + "learning_rate": 2.2708083952929634e-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": 0.009220260940492154, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.004256098275072872, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.022696620132774115, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.02096653329208493, + "weighted_rkd_loss": 0.011348310066387058 + }, + { + "epoch": 3.053691275167785, + "flow_matching_loss": 0.022351188492029904, + "grad_norm": 0.2090246081352234, + "learning_rate": 2.2622808582104594e-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": 0.00882717128843069, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0034880529856309295, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.02114239577203989, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.022351188492029904, + "weighted_rkd_loss": 0.010571197886019946 + }, + { + "epoch": 3.0704697986577183, + "flow_matching_loss": 0.020570892561227083, + "grad_norm": 0.18636099994182587, + "learning_rate": 2.2537199764937703e-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": 0.009009950887411832, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.004132049717009067, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.02215195130556822, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.020570892561227083, + "weighted_rkd_loss": 0.01107597565278411 + }, + { + "epoch": 3.087248322147651, + "flow_matching_loss": 0.02188956905156374, + "grad_norm": 0.18344570696353912, + "learning_rate": 2.245126124623595e-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": 0.008544707810506225, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.00419144555926323, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.021280860900878905, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.02188956905156374, + "weighted_rkd_loss": 0.010640430450439452 + }, + { + "epoch": 3.1040268456375837, + "flow_matching_loss": 0.022426276095211507, + "grad_norm": 0.2003953754901886, + "learning_rate": 2.2364996785228547e-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": 0.008748158859089018, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.003919238899834454, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.02141555678099394, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.022426276095211507, + "weighted_rkd_loss": 0.01070777839049697 + }, + { + "epoch": 3.120805369127517, + "flow_matching_loss": 0.02033795341849327, + "grad_norm": 0.22842928767204285, + "learning_rate": 2.227841015540248e-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": 0.007575100101530552, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0034250223077833654, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01857522251084447, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.02033795341849327, + "weighted_rkd_loss": 0.009287611255422235 + }, + { + "epoch": 3.1375838926174495, + "flow_matching_loss": 0.020844181906431912, + "grad_norm": 0.2263597548007965, + "learning_rate": 2.2191505144337433e-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": 0.007685949653387069, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.003535176650620997, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.018907075840979816, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.020844181906431912, + "weighted_rkd_loss": 0.009453537920489908 + }, + { + "epoch": 3.1543624161073827, + "flow_matching_loss": 0.019808660540729762, + "grad_norm": 0.18585313856601715, + "learning_rate": 2.210428555354013e-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": 0.009581155283376575, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.004591489001177252, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.023753799498081207, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.019808660540729762, + "weighted_rkd_loss": 0.011876899749040604 + }, + { + "epoch": 3.1711409395973154, + "flow_matching_loss": 0.020991289615631105, + "grad_norm": 0.22622573375701904, + "learning_rate": 2.201675519827802e-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": 0.009574074717238545, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.004285450396127999, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.023433599807322025, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.020991289615631105, + "weighted_rkd_loss": 0.011716799903661012 + }, + { + "epoch": 3.1879194630872485, + "flow_matching_loss": 0.02278821486979723, + "grad_norm": 0.22967985272407532, + "learning_rate": 2.1928917907412403e-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": 0.009182234713807701, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0036421044263988734, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.022006573714315893, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.02278821486979723, + "weighted_rkd_loss": 0.011003286857157946 + }, + { + "epoch": 3.204697986577181, + "flow_matching_loss": 0.021890855114907026, + "grad_norm": 0.1822008639574051, + "learning_rate": 2.1840777523230937e-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": 0.008652022993192076, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.004090635036118328, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.02139468090608716, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.021890855114907026, + "weighted_rkd_loss": 0.01069734045304358 + }, + { + "epoch": 3.221476510067114, + "flow_matching_loss": 0.02066585961729288, + "grad_norm": 0.1782473474740982, + "learning_rate": 2.1752337901279552e-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": 0.00783657394349575, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.003442372544668615, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.019115520268678667, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.02066585961729288, + "weighted_rkd_loss": 0.009557760134339333 + }, + { + "epoch": 3.238255033557047, + "flow_matching_loss": 0.020921053178608418, + "grad_norm": 0.21969790756702423, + "learning_rate": 2.166360291019383e-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": 0.009407800249755382, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0038952767383307217, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.02271087756380439, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.020921053178608418, + "weighted_rkd_loss": 0.011355438781902195 + }, + { + "epoch": 3.2550335570469797, + "flow_matching_loss": 0.01944717802107334, + "grad_norm": 0.19404731690883636, + "learning_rate": 2.157457643152973e-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": 0.00837903949432075, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0037874118657782673, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.020545491110533477, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.01944717802107334, + "weighted_rkd_loss": 0.010272745555266739 + }, + { + "epoch": 3.271812080536913, + "flow_matching_loss": 0.0224097715690732, + "grad_norm": 0.17556366324424744, + "learning_rate": 2.148526235959385e-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": 0.008008029777556658, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0038434152025729417, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01985947471112013, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.0224097715690732, + "weighted_rkd_loss": 0.009929737355560065 + }, + { + "epoch": 3.2885906040268456, + "flow_matching_loss": 0.019902830384671687, + "grad_norm": 0.2150319367647171, + "learning_rate": 2.1395664601273016e-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": 0.008881382131949066, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0041137875756248835, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.021876551769673826, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.019902830384671687, + "weighted_rkd_loss": 0.010938275884836913 + }, + { + "epoch": 3.3053691275167787, + "flow_matching_loss": 0.020638061966747046, + "grad_norm": 0.20082297921180725, + "learning_rate": 2.130578707586345e-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": 0.009990708529949188, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.004830724210478365, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.024812141060829164, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.020638061966747046, + "weighted_rkd_loss": 0.012406070530414582 + }, + { + "epoch": 3.3221476510067114, + "flow_matching_loss": 0.021972791384905578, + "grad_norm": 0.36457595229148865, + "learning_rate": 2.1215633714899264e-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": 0.008448011986911298, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.004079858679324389, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.020975882653146982, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.021972791384905578, + "weighted_rkd_loss": 0.010487941326573491 + }, + { + "epoch": 3.3389261744966445, + "flow_matching_loss": 0.020128148049116133, + "grad_norm": 0.2117772102355957, + "learning_rate": 2.112520846198053e-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": 0.009189584525302052, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.003930398868396879, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.022309568151831626, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.020128148049116133, + "weighted_rkd_loss": 0.011154784075915813 + }, + { + "epoch": 3.3557046979865772, + "flow_matching_loss": 0.02076246850192547, + "grad_norm": 0.22179049253463745, + "learning_rate": 2.103451527260074e-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": 0.0077252968214452265, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0034601765451952814, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.018910769931972028, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.02076246850192547, + "weighted_rkd_loss": 0.009455384965986014 + }, + { + "epoch": 3.37248322147651, + "flow_matching_loss": 0.019000662211328744, + "grad_norm": 0.22441506385803223, + "learning_rate": 2.094355811397381e-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": 0.007931182719767093, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0037250424269586802, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01958740781992674, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.019000662211328744, + "weighted_rkd_loss": 0.00979370390996337 + }, + { + "epoch": 3.389261744966443, + "flow_matching_loss": 0.02324447510764003, + "grad_norm": 0.18134300410747528, + "learning_rate": 2.0852340964860522e-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": 0.007788776373490691, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0033203586237505077, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.018897911440581082, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.02324447510764003, + "weighted_rkd_loss": 0.009448955720290541 + }, + { + "epoch": 3.4060402684563758, + "flow_matching_loss": 0.02150107007473707, + "grad_norm": 0.20962178707122803, + "learning_rate": 2.076086781539447e-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": 0.0074673259630799295, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0032554172910749914, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.018190069310367107, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.02150107007473707, + "weighted_rkd_loss": 0.009095034655183553 + }, + { + "epoch": 3.422818791946309, + "flow_matching_loss": 0.018979523424059154, + "grad_norm": 0.2038407325744629, + "learning_rate": 2.066914266690755e-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": 0.008069745684042572, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0033081255503930153, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.019447616953402757, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.018979523424059154, + "weighted_rkd_loss": 0.009723808476701378 + }, + { + "epoch": 3.4395973154362416, + "flow_matching_loss": 0.018519976176321507, + "grad_norm": 0.2189130187034607, + "learning_rate": 2.0577169531754906e-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": 0.008641075901687145, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.003642701031640172, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.02092485288158059, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.018519976176321507, + "weighted_rkd_loss": 0.010462426440790295 + }, + { + "epoch": 3.4563758389261743, + "flow_matching_loss": 0.023320261761546134, + "grad_norm": 0.19270910322666168, + "learning_rate": 2.048495243313942e-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": 0.006468802364543081, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.003018935560248792, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.015956540312618017, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.023320261761546134, + "weighted_rkd_loss": 0.007978270156309008 + }, + { + "epoch": 3.4731543624161074, + "flow_matching_loss": 0.01975588295608759, + "grad_norm": 0.1822284758090973, + "learning_rate": 2.039249540493572e-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": 0.008690228406339883, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.00411738664843142, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.02149784378707409, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.01975588295608759, + "weighted_rkd_loss": 0.010748921893537045 + }, + { + "epoch": 3.48993288590604, + "flow_matching_loss": 0.01874214904382825, + "grad_norm": 0.1515883505344391, + "learning_rate": 2.0299802491513755e-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": 0.008525742217898369, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0038174554938450456, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.020868940092623234, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.01874214904382825, + "weighted_rkd_loss": 0.010434470046311617 + }, + { + "epoch": 3.5067114093959733, + "flow_matching_loss": 0.018373808544129133, + "grad_norm": 0.22102390229701996, + "learning_rate": 2.0206877747561828e-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": 0.007662647403776646, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0035474081989377735, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.018872702866792677, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.018373808544129133, + "weighted_rkd_loss": 0.009436351433396339 + }, + { + "epoch": 3.523489932885906, + "flow_matching_loss": 0.017856397293508054, + "grad_norm": 0.15369464457035065, + "learning_rate": 2.011372523790927e-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": 0.008157176710665227, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0034173124935477973, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.019731666054576637, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.017856397293508054, + "weighted_rkd_loss": 0.009865833027288318 + }, + { + "epoch": 3.540268456375839, + "flow_matching_loss": 0.01998805645853281, + "grad_norm": 0.1758951097726822, + "learning_rate": 2.0020349037348643e-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": 0.006989141134545207, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0031998127698898314, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.017178095038980247, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.01998805645853281, + "weighted_rkd_loss": 0.008589047519490123 + }, + { + "epoch": 3.557046979865772, + "flow_matching_loss": 0.018932559806853533, + "grad_norm": 0.17690835893154144, + "learning_rate": 1.9926753230457443e-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": 0.007896357215940952, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0036027796799317003, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.0193954941816628, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.018932559806853533, + "weighted_rkd_loss": 0.0096977470908314 + }, + { + "epoch": 3.573825503355705, + "flow_matching_loss": 0.023601194098591804, + "grad_norm": 0.19478943943977356, + "learning_rate": 1.983294191141948e-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": 0.009023548383265734, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.004056708514690399, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.02210380509495735, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.023601194098591804, + "weighted_rkd_loss": 0.011051902547478676 + }, + { + "epoch": 3.5906040268456376, + "flow_matching_loss": 0.020795693341642617, + "grad_norm": 0.15999910235404968, + "learning_rate": 1.973891918384575e-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": 0.008007427817210555, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.003208912722766399, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01922376835718751, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.020795693341642617, + "weighted_rkd_loss": 0.009611884178593755 + }, + { + "epoch": 3.6073825503355703, + "flow_matching_loss": 0.021222106367349624, + "grad_norm": 0.21778394281864166, + "learning_rate": 1.9644689160594945e-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": 0.00793103752657771, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.003410175652243197, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01927225077524781, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.021222106367349624, + "weighted_rkd_loss": 0.009636125387623906 + }, + { + "epoch": 3.6241610738255035, + "flow_matching_loss": 0.017535629961639643, + "grad_norm": 0.21843908727169037, + "learning_rate": 1.9550255963593553e-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": 0.007544249249622226, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0029765219427645207, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.018065020348876713, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.017535629961639643, + "weighted_rkd_loss": 0.009032510174438357 + }, + { + "epoch": 3.640939597315436, + "flow_matching_loss": 0.019194155745208264, + "grad_norm": 0.20880937576293945, + "learning_rate": 1.9455623723655524e-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": 0.00752829727716744, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0031787700252607467, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.018235364742577077, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.019194155745208264, + "weighted_rkd_loss": 0.009117682371288538 + }, + { + "epoch": 3.6577181208053693, + "flow_matching_loss": 0.019468952622264623, + "grad_norm": 0.18948209285736084, + "learning_rate": 1.9360796580301606e-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": 0.007364970911294222, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0032614518189802767, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.017991393711417912, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.019468952622264623, + "weighted_rkd_loss": 0.008995696855708956 + }, + { + "epoch": 3.674496644295302, + "flow_matching_loss": 0.022077844850718974, + "grad_norm": 0.16150549054145813, + "learning_rate": 1.9265778681578245e-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": 0.00765900406986475, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0032386504113674163, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01855665845796466, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.022077844850718974, + "weighted_rkd_loss": 0.00927832922898233 + }, + { + "epoch": 3.6912751677852347, + "flow_matching_loss": 0.019230139348655938, + "grad_norm": 0.25615647435188293, + "learning_rate": 1.9170574183876143e-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": 0.007380595942959189, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0032563595566898586, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01801755167543888, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.019230139348655938, + "weighted_rkd_loss": 0.00900877583771944 + }, + { + "epoch": 3.708053691275168, + "flow_matching_loss": 0.017652079928666353, + "grad_norm": 0.1817786544561386, + "learning_rate": 1.9075187251748452e-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": 0.007373360265046358, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0032471208134666084, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.017993841134011746, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.017652079928666353, + "weighted_rkd_loss": 0.008996920567005873 + }, + { + "epoch": 3.7248322147651005, + "flow_matching_loss": 0.01961641861125827, + "grad_norm": 0.17127038538455963, + "learning_rate": 1.8979622057728596e-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": 0.00780273643322289, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.003444134397432208, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.019049607403576373, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.01961641861125827, + "weighted_rkd_loss": 0.009524803701788187 + }, + { + "epoch": 3.7416107382550337, + "flow_matching_loss": 0.019033155217766762, + "grad_norm": 0.17839007079601288, + "learning_rate": 1.8883882782147748e-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": 0.00856158221140504, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.003942649811506271, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.02106581414118409, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.019033155217766762, + "weighted_rkd_loss": 0.010532907070592045 + }, + { + "epoch": 3.7583892617449663, + "flow_matching_loss": 0.019040965475142002, + "grad_norm": 0.1941649317741394, + "learning_rate": 1.8787973612951967e-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": 0.007480817148461938, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0031261443509720266, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01808777851983905, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.019040965475142002, + "weighted_rkd_loss": 0.009043889259919525 + }, + { + "epoch": 3.7751677852348995, + "flow_matching_loss": 0.019748223666101693, + "grad_norm": 0.1636851280927658, + "learning_rate": 1.8691898745519028e-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": 0.008198972046375274, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.003771033138036728, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.02016897704452276, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.019748223666101693, + "weighted_rkd_loss": 0.01008448852226138 + }, + { + "epoch": 3.791946308724832, + "flow_matching_loss": 0.022198669519275425, + "grad_norm": 0.22628392279148102, + "learning_rate": 1.8595662382474863e-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": 0.007077188324183225, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0031255503417924047, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.017279926873743533, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.022198669519275425, + "weighted_rkd_loss": 0.008639963436871767 + }, + { + "epoch": 3.8087248322147653, + "flow_matching_loss": 0.016530673112720252, + "grad_norm": 0.21448732912540436, + "learning_rate": 1.8499268733509766e-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": 0.007735572382807731, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0035640710266306996, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.019035216141492128, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.016530673112720252, + "weighted_rkd_loss": 0.009517608070746064 + }, + { + "epoch": 3.825503355704698, + "flow_matching_loss": 0.021276569552719594, + "grad_norm": 0.14651033282279968, + "learning_rate": 1.84027220151942e-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": 0.00765763008967042, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0031805066973902285, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01849576672539115, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.021276569552719594, + "weighted_rkd_loss": 0.009247883362695575 + }, + { + "epoch": 3.8422818791946307, + "flow_matching_loss": 0.021208910923451185, + "grad_norm": 0.1972508579492569, + "learning_rate": 1.83060264507944e-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": 0.007044157804921269, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0029331859084777536, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.017021501343697308, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.021208910923451185, + "weighted_rkd_loss": 0.008510750671848654 + }, + { + "epoch": 3.859060402684564, + "flow_matching_loss": 0.018153464980423452, + "grad_norm": 0.1617676019668579, + "learning_rate": 1.82091862700876e-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": 0.008048683824017643, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.003274902165867388, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.019372269883751868, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.018153464980423452, + "weighted_rkd_loss": 0.009686134941875934 + }, + { + "epoch": 3.8758389261744965, + "flow_matching_loss": 0.019316170644015074, + "grad_norm": 0.20088458061218262, + "learning_rate": 1.8112205709177023e-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": 0.00764038790948689, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0031374069745652377, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01841818243265152, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.019316170644015074, + "weighted_rkd_loss": 0.00920909121632576 + }, + { + "epoch": 3.8926174496644297, + "flow_matching_loss": 0.01878544017672539, + "grad_norm": 0.17331726849079132, + "learning_rate": 1.8015089010306585e-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": 0.006989720230922103, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0030413487227633597, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.0170207891613245, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.01878544017672539, + "weighted_rkd_loss": 0.00851039458066225 + }, + { + "epoch": 3.9093959731543624, + "flow_matching_loss": 0.018327188584953545, + "grad_norm": 0.18315960466861725, + "learning_rate": 1.7917840421675298e-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": 0.006432220642454922, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0026728694792836905, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.015537310857325792, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.018327188584953545, + "weighted_rkd_loss": 0.007768655428662896 + }, + { + "epoch": 3.926174496644295, + "flow_matching_loss": 0.017242155969142914, + "grad_norm": 0.17186187207698822, + "learning_rate": 1.7820464197251474e-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": 0.006151645421050489, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0027259318390861153, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.015029222704470158, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.017242155969142914, + "weighted_rkd_loss": 0.007514611352235079 + }, + { + "epoch": 3.942953020134228, + "flow_matching_loss": 0.02237462541088462, + "grad_norm": 0.24216608703136444, + "learning_rate": 1.7722964596586627e-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": 0.007923036813735962, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.003261329559609294, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01910740341991186, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.02237462541088462, + "weighted_rkd_loss": 0.00955370170995593 + }, + { + "epoch": 3.959731543624161, + "flow_matching_loss": 0.018751779105514288, + "grad_norm": 0.23497825860977173, + "learning_rate": 1.7625345884629144e-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": 0.007246668357402087, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0031278782756999136, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01762121506035328, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.018751779105514288, + "weighted_rkd_loss": 0.00881060753017664 + }, + { + "epoch": 3.976510067114094, + "flow_matching_loss": 0.01858077123761177, + "grad_norm": 0.20044608414173126, + "learning_rate": 1.7527612331537722e-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": 0.006888248072937131, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.002850493707228452, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.016626989655196666, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.01858077123761177, + "weighted_rkd_loss": 0.008313494827598333 + }, + { + "epoch": 3.9932885906040267, + "flow_matching_loss": 0.019617028534412384, + "grad_norm": 0.19277773797512054, + "learning_rate": 1.7429768212494593e-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": 0.007408017152920365, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.00341418138705194, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.0182302157394588, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.019617028534412384, + "weighted_rkd_loss": 0.0091151078697294 + }, + { + "epoch": 4.010067114093959, + "flow_matching_loss": 0.01653716042637825, + "grad_norm": 0.16038158535957336, + "learning_rate": 1.7331817807518497e-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": 0.006630103243514896, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.003022903867531568, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01628311024978757, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.01653716042637825, + "weighted_rkd_loss": 0.008141555124893785 + }, + { + "epoch": 4.026845637583893, + "flow_matching_loss": 0.018175854813307524, + "grad_norm": 0.19850464165210724, + "learning_rate": 1.723376540127745e-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": 0.00760634196922183, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0034310010029003023, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.0186436852440238, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.018175854813307524, + "weighted_rkd_loss": 0.0093218426220119 + }, + { + "epoch": 4.043624161073826, + "flow_matching_loss": 0.019848077557981014, + "grad_norm": 0.16854286193847656, + "learning_rate": 1.7135615282901357e-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": 0.008194463048130273, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.00404913816601038, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.020438064355403185, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.019848077557981014, + "weighted_rkd_loss": 0.010219032177701592 + }, + { + "epoch": 4.060402684563758, + "flow_matching_loss": 0.020925424993038177, + "grad_norm": 0.20957912504673004, + "learning_rate": 1.7037371745794358e-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": 0.006662683980539441, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0029504644102416933, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.016275832336395978, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.020925424993038177, + "weighted_rkd_loss": 0.008137916168197989 + }, + { + "epoch": 4.077181208053691, + "flow_matching_loss": 0.019630491454154252, + "grad_norm": 0.1673985868692398, + "learning_rate": 1.6939039087447026e-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": 0.006573117896914482, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.002810140186920762, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01595637621358037, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.019630491454154252, + "weighted_rkd_loss": 0.007978188106790185 + }, + { + "epoch": 4.093959731543624, + "flow_matching_loss": 0.021586596686393022, + "grad_norm": 0.201747328042984, + "learning_rate": 1.6840621609248397e-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": 0.007233408000320196, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.003126069717109203, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01759288590401411, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.021586596686393022, + "weighted_rkd_loss": 0.008796442952007055 + }, + { + "epoch": 4.110738255033557, + "flow_matching_loss": 0.018460760451853277, + "grad_norm": 0.2036897987127304, + "learning_rate": 1.6742123616297794e-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": 0.007287621404975653, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.003143525216728449, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.017718768026679755, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.018460760451853277, + "weighted_rkd_loss": 0.008859384013339877 + }, + { + "epoch": 4.12751677852349, + "flow_matching_loss": 0.017445159237831832, + "grad_norm": 0.16198286414146423, + "learning_rate": 1.664354941721651e-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": 0.006410072930157185, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.003017538005951792, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.0158376838080585, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.017445159237831832, + "weighted_rkd_loss": 0.00791884190402925 + }, + { + "epoch": 4.144295302013423, + "flow_matching_loss": 0.016915331315249204, + "grad_norm": 0.16699963808059692, + "learning_rate": 1.6544903323959355e-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": 0.005673299496993423, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0020931367063894866, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.013439735677093268, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.016915331315249204, + "weighted_rkd_loss": 0.006719867838546634 + }, + { + "epoch": 4.1610738255033555, + "flow_matching_loss": 0.020208620838820933, + "grad_norm": 0.215232715010643, + "learning_rate": 1.6446189651626023e-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": 0.006692374125123024, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0026895080460235476, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01607425631955266, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.020208620838820933, + "weighted_rkd_loss": 0.00803712815977633 + }, + { + "epoch": 4.177852348993288, + "flow_matching_loss": 0.020926382672041655, + "grad_norm": 0.22336435317993164, + "learning_rate": 1.6347412718272316e-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": 0.007076423778198659, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.003159638389479369, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01731248591095209, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.020926382672041655, + "weighted_rkd_loss": 0.008656242955476045 + }, + { + "epoch": 4.194630872483222, + "flow_matching_loss": 0.016524071246385573, + "grad_norm": 0.1518680602312088, + "learning_rate": 1.624857684472131e-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": 0.006714969547465444, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.003003756864927709, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01643369598314166, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.016524071246385573, + "weighted_rkd_loss": 0.00821684799157083 + }, + { + "epoch": 4.2114093959731544, + "flow_matching_loss": 0.018874516151845454, + "grad_norm": 0.17943860590457916, + "learning_rate": 1.6149686354374283e-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": 0.006851326674222946, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.002800596947781742, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.0165032503195107, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.018874516151845454, + "weighted_rkd_loss": 0.00825162515975535 + }, + { + "epoch": 4.228187919463087, + "flow_matching_loss": 0.016614831425249575, + "grad_norm": 0.25215211510658264, + "learning_rate": 1.605074557302165e-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": 0.006883031688630581, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0028047516010701657, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01657081479206681, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.016614831425249575, + "weighted_rkd_loss": 0.008285407396033406 + }, + { + "epoch": 4.24496644295302, + "flow_matching_loss": 0.01997323101386428, + "grad_norm": 0.20624391734600067, + "learning_rate": 1.59517588286537e-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": 0.006685473956167698, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.00293945602606982, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01631040396168828, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.01997323101386428, + "weighted_rkd_loss": 0.00815520198084414 + }, + { + "epoch": 4.261744966442953, + "flow_matching_loss": 0.02013480234891176, + "grad_norm": 0.1708594560623169, + "learning_rate": 1.5852730451271324e-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": 0.005825302423909306, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.002814886881969869, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.014465491846203804, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.02013480234891176, + "weighted_rkd_loss": 0.007232745923101902 + }, + { + "epoch": 4.278523489932886, + "flow_matching_loss": 0.017857098765671254, + "grad_norm": 0.16984955966472626, + "learning_rate": 1.5753664772696546e-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": 0.006671967078000307, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0027174580260179936, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01606139214709401, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.017857098765671254, + "weighted_rkd_loss": 0.008030696073547006 + }, + { + "epoch": 4.295302013422819, + "flow_matching_loss": 0.017334267776459456, + "grad_norm": 0.15617665648460388, + "learning_rate": 1.565456612638307e-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": 0.006426696386188269, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.002855884563177824, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.015709277335554363, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.017334267776459456, + "weighted_rkd_loss": 0.007854638667777181 + }, + { + "epoch": 4.3120805369127515, + "flow_matching_loss": 0.019305859599262475, + "grad_norm": 0.16668863594532013, + "learning_rate": 1.5555438847226732e-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": 0.00716956895776093, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0026791229262016713, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.017018260527402164, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.019305859599262475, + "weighted_rkd_loss": 0.008509130263701082 + }, + { + "epoch": 4.328859060402684, + "flow_matching_loss": 0.019607679545879365, + "grad_norm": 0.1785610169172287, + "learning_rate": 1.545628727137584e-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": 0.009049271885305642, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0035267358762212098, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.021625279542058708, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.019607679545879365, + "weighted_rkd_loss": 0.010812639771029354 + }, + { + "epoch": 4.345637583892618, + "flow_matching_loss": 0.018800497893244027, + "grad_norm": 0.1957562118768692, + "learning_rate": 1.535711573604153e-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": 0.00803911597467959, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.003370839380659163, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.019449071399867533, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.018800497893244027, + "weighted_rkd_loss": 0.009724535699933767 + }, + { + "epoch": 4.3624161073825505, + "flow_matching_loss": 0.017709657829254866, + "grad_norm": 0.2177783101797104, + "learning_rate": 1.5257928579308032e-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": 0.006389908026903868, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0026443781214766206, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.015424194186925888, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.017709657829254866, + "weighted_rkd_loss": 0.007712097093462944 + }, + { + "epoch": 4.379194630872483, + "flow_matching_loss": 0.01953520327806473, + "grad_norm": 0.22895203530788422, + "learning_rate": 1.5158730139942902e-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": 0.008311857981607319, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.003941608686000108, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.02056532446295023, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.01953520327806473, + "weighted_rkd_loss": 0.010282662231475115 + }, + { + "epoch": 4.395973154362416, + "flow_matching_loss": 0.020596037432551385, + "grad_norm": 0.13484011590480804, + "learning_rate": 1.5059524757207239e-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": 0.005220752139575779, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0022500797058455646, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01269158385694027, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.020596037432551385, + "weighted_rkd_loss": 0.006345791928470135 + }, + { + "epoch": 4.412751677852349, + "flow_matching_loss": 0.018211192917078733, + "grad_norm": 0.17946451902389526, + "learning_rate": 1.4960316770665866e-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": 0.006453538197092712, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0030796703533269463, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.015986746829003095, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.018211192917078733, + "weighted_rkd_loss": 0.007993373414501547 + }, + { + "epoch": 4.429530201342282, + "flow_matching_loss": 0.0171947973780334, + "grad_norm": 0.1605665385723114, + "learning_rate": 1.4861110519997495e-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": 0.007121807662770152, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0029179507866501806, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.017161566112190486, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.0171947973780334, + "weighted_rkd_loss": 0.008580783056095243 + }, + { + "epoch": 4.446308724832215, + "flow_matching_loss": 0.017047136928886174, + "grad_norm": 0.21463190019130707, + "learning_rate": 1.4761910344804913e-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": 0.006163895828649402, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0027821387513540683, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01510993055999279, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.017047136928886174, + "weighted_rkd_loss": 0.007554965279996395 + }, + { + "epoch": 4.4630872483221475, + "flow_matching_loss": 0.019098941888660192, + "grad_norm": 0.13163383305072784, + "learning_rate": 1.466272058442516e-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": 0.005966320005245507, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0023630001582205296, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.0142956402618438, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.019098941888660192, + "weighted_rkd_loss": 0.0071478201309219 + }, + { + "epoch": 4.47986577181208, + "flow_matching_loss": 0.01778299594298005, + "grad_norm": 0.2292172610759735, + "learning_rate": 1.4563545577739676e-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": 0.006542552821338176, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0030771528254263103, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01616225838661194, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.01778299594298005, + "weighted_rkd_loss": 0.00808112919330597 + }, + { + "epoch": 4.496644295302014, + "flow_matching_loss": 0.015495003666728735, + "grad_norm": 0.17960447072982788, + "learning_rate": 1.4464389662984543e-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": 0.007274207659065723, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.002846747334115207, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.017395162582397462, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.015495003666728735, + "weighted_rkd_loss": 0.008697581291198731 + }, + { + "epoch": 4.5134228187919465, + "flow_matching_loss": 0.016512521728873252, + "grad_norm": 0.18700137734413147, + "learning_rate": 1.4365257177560684e-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": 0.006523366179317236, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0025913154473528265, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.015638047736138104, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.016512521728873252, + "weighted_rkd_loss": 0.007819023868069052 + }, + { + "epoch": 4.530201342281879, + "flow_matching_loss": 0.018928809091448784, + "grad_norm": 0.1533912718296051, + "learning_rate": 1.426615245784416e-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": 0.006431228248402476, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0026095522684045137, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.015472008753567934, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.018928809091448784, + "weighted_rkd_loss": 0.007736004376783967 + }, + { + "epoch": 4.546979865771812, + "flow_matching_loss": 0.01595159135758877, + "grad_norm": 0.13736169040203094, + "learning_rate": 1.4167079838996468e-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": 0.006889982311986387, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0030951215769164262, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01687508588656783, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.01595159135758877, + "weighted_rkd_loss": 0.008437542943283916 + }, + { + "epoch": 4.563758389261745, + "flow_matching_loss": 0.017109850142151117, + "grad_norm": 0.1820959895849228, + "learning_rate": 1.4068043654774897e-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": 0.006259463052265346, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0027393482509069146, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01525827432051301, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.017109850142151117, + "weighted_rkd_loss": 0.007629137160256505 + }, + { + "epoch": 4.580536912751678, + "flow_matching_loss": 0.018156492337584495, + "grad_norm": 0.182673841714859, + "learning_rate": 1.3969048237342992e-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": 0.0066155354026705025, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0028239200590178372, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01605499079450965, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.018156492337584495, + "weighted_rkd_loss": 0.008027495397254825 + }, + { + "epoch": 4.597315436241611, + "flow_matching_loss": 0.01978831607848406, + "grad_norm": 0.20100091397762299, + "learning_rate": 1.3870097917081011e-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": 0.0070770631544291975, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.002857843821402639, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.017011970281600952, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.01978831607848406, + "weighted_rkd_loss": 0.008505985140800476 + }, + { + "epoch": 4.614093959731544, + "flow_matching_loss": 0.0186252910643816, + "grad_norm": 0.18295343220233917, + "learning_rate": 1.3771197022396508e-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": 0.006896005268208683, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.002676388737745583, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.016468399157747626, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.0186252910643816, + "weighted_rkd_loss": 0.008234199578873813 + }, + { + "epoch": 4.630872483221476, + "flow_matching_loss": 0.017001602426171304, + "grad_norm": 0.18725721538066864, + "learning_rate": 1.3672349879535015e-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": 0.0075272177811712025, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.00305442416574806, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01810885975137353, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.017001602426171304, + "weighted_rkd_loss": 0.009054429875686765 + }, + { + "epoch": 4.64765100671141, + "flow_matching_loss": 0.01720100613310933, + "grad_norm": 0.15701332688331604, + "learning_rate": 1.3573560812390781e-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": 0.006909766187891364, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0026912813424132763, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.016510813497006893, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.01720100613310933, + "weighted_rkd_loss": 0.008255406748503447 + }, + { + "epoch": 4.6644295302013425, + "flow_matching_loss": 0.017884473502635955, + "grad_norm": 0.14792506396770477, + "learning_rate": 1.3474834142317635e-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": 0.006335597159340978, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.002531678543891758, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.015202872920781374, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.017884473502635955, + "weighted_rkd_loss": 0.007601436460390687 + }, + { + "epoch": 4.681208053691275, + "flow_matching_loss": 0.01976926876232028, + "grad_norm": 0.17041365802288055, + "learning_rate": 1.3376174187939955e-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": 0.0057981258258223535, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.002652096108067781, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01424834793433547, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.01976926876232028, + "weighted_rkd_loss": 0.007124173967167735 + }, + { + "epoch": 4.697986577181208, + "flow_matching_loss": 0.019026348367333412, + "grad_norm": 0.15966573357582092, + "learning_rate": 1.3277585264963761e-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": 0.0060622354969382284, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0023348112939856947, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.014459282439202071, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.019026348367333412, + "weighted_rkd_loss": 0.0072296412196010355 + }, + { + "epoch": 4.714765100671141, + "flow_matching_loss": 0.017377084400504828, + "grad_norm": 0.22258582711219788, + "learning_rate": 1.3179071685987928e-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": 0.0056399278808385135, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.002916206861846149, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.014196062553673982, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.017377084400504828, + "weighted_rkd_loss": 0.007098031276836991 + }, + { + "epoch": 4.731543624161074, + "flow_matching_loss": 0.017920501250773668, + "grad_norm": 0.2010204941034317, + "learning_rate": 1.308063776031554e-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": 0.005940242158249021, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.002402682788670063, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.014283167012035847, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.017920501250773668, + "weighted_rkd_loss": 0.007141583506017924 + }, + { + "epoch": 4.748322147651007, + "flow_matching_loss": 0.01623355569317937, + "grad_norm": 0.17003734409809113, + "learning_rate": 1.2982287793765389e-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": 0.005624843528494239, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0024171904427930714, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.013666877429932356, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.01623355569317937, + "weighted_rkd_loss": 0.006833438714966178 + }, + { + "epoch": 4.76510067114094, + "flow_matching_loss": 0.016382606141269206, + "grad_norm": 0.16437426209449768, + "learning_rate": 1.2884026088483616e-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": 0.005247332667931914, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0022283923462964594, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.012723057717084884, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.016382606141269206, + "weighted_rkd_loss": 0.006361528858542442 + }, + { + "epoch": 4.781879194630872, + "flow_matching_loss": 0.020086504332721234, + "grad_norm": 0.20742954313755035, + "learning_rate": 1.2785856942755543e-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": 0.006500152428634465, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0027116223354823887, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.015711927181109787, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.020086504332721234, + "weighted_rkd_loss": 0.007855963590554894 + }, + { + "epoch": 4.798657718120805, + "flow_matching_loss": 0.020468197017908096, + "grad_norm": 0.18006539344787598, + "learning_rate": 1.2687784650817616e-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": 0.006521564652211964, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0030275082564912735, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.016070637572556733, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.020468197017908096, + "weighted_rkd_loss": 0.008035318786278367 + }, + { + "epoch": 4.815436241610739, + "flow_matching_loss": 0.020298132486641408, + "grad_norm": 0.21230530738830566, + "learning_rate": 1.2589813502669603e-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": 0.006211647461168468, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.002518127963412553, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.014941422827541828, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.020298132486641408, + "weighted_rkd_loss": 0.007470711413770914 + }, + { + "epoch": 4.832214765100671, + "flow_matching_loss": 0.018210323061794043, + "grad_norm": 0.17627596855163574, + "learning_rate": 1.2491947783886907e-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": 0.004903101362287998, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0021681559388525783, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.011974358558654785, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.018210323061794043, + "weighted_rkd_loss": 0.005987179279327392 + }, + { + "epoch": 4.848993288590604, + "flow_matching_loss": 0.016133444383740425, + "grad_norm": 0.13107307255268097, + "learning_rate": 1.2394191775433106e-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": 0.005635178997181356, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.002382622950244695, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.013652980979532003, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.016133444383740425, + "weighted_rkd_loss": 0.006826490489766001 + }, + { + "epoch": 4.865771812080537, + "flow_matching_loss": 0.016500497329980136, + "grad_norm": 0.1938703954219818, + "learning_rate": 1.2296549753472696e-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": 0.005910794297233224, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.002559423819184303, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.014381012320518494, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.016500497329980136, + "weighted_rkd_loss": 0.007190506160259247 + }, + { + "epoch": 4.882550335570469, + "flow_matching_loss": 0.015015983954071999, + "grad_norm": 0.18545067310333252, + "learning_rate": 1.2199025989184032e-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": 0.005890693236142397, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0025859495741315185, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.014367335941642523, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.015015983954071999, + "weighted_rkd_loss": 0.007183667970821261 + }, + { + "epoch": 4.899328859060403, + "flow_matching_loss": 0.015698899049311876, + "grad_norm": 0.20254436135292053, + "learning_rate": 1.21016247485725e-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": 0.0064388867001980545, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0024498644983395935, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01532763782888651, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.015698899049311876, + "weighted_rkd_loss": 0.007663818914443255 + }, + { + "epoch": 4.916107382550336, + "flow_matching_loss": 0.016660439129918814, + "grad_norm": 0.18414856493473053, + "learning_rate": 1.2004350292283897e-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": 0.006194327911362052, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0027924057794734836, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.015181061439216137, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.016660439129918814, + "weighted_rkd_loss": 0.007590530719608068 + }, + { + "epoch": 4.932885906040268, + "flow_matching_loss": 0.01687433701008558, + "grad_norm": 0.16476738452911377, + "learning_rate": 1.1907206875418068e-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": 0.006046470603905618, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0024416908156126736, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.014534632116556168, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.01687433701008558, + "weighted_rkd_loss": 0.007267316058278084 + }, + { + "epoch": 4.949664429530201, + "flow_matching_loss": 0.016659815981984138, + "grad_norm": 0.15063215792179108, + "learning_rate": 1.1810198747342769e-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": 0.005068508721888065, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0021296532708220185, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.012266670726239682, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.016659815981984138, + "weighted_rkd_loss": 0.006133335363119841 + }, + { + "epoch": 4.966442953020135, + "flow_matching_loss": 0.014483388233929872, + "grad_norm": 0.18998196721076965, + "learning_rate": 1.1713330151507786e-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": 0.006280959723517299, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0027005726704373957, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.015262492187321185, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.014483388233929872, + "weighted_rkd_loss": 0.007631246093660593 + }, + { + "epoch": 4.983221476510067, + "flow_matching_loss": 0.016974644828587772, + "grad_norm": 0.1717025190591812, + "learning_rate": 1.161660532525931e-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": 0.006291400594636798, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0028575255651958286, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.015440326556563378, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.016974644828587772, + "weighted_rkd_loss": 0.007720163278281689 + }, + { + "epoch": 5.0, + "flow_matching_loss": 0.015258622635155917, + "grad_norm": 0.16946367919445038, + "learning_rate": 1.1520028499654584e-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": 0.00629035490565002, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.002330243296455592, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.014910953119397163, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.015258622635155917, + "weighted_rkd_loss": 0.007455476559698582 + }, + { + "epoch": 5.016778523489933, + "flow_matching_loss": 0.017898726742714643, + "grad_norm": 0.14455287158489227, + "learning_rate": 1.1423603899276833e-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": 0.0053604997927322985, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.002361709764227271, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01308270925655961, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.017898726742714643, + "weighted_rkd_loss": 0.006541354628279805 + }, + { + "epoch": 5.033557046979865, + "flow_matching_loss": 0.01671588756144047, + "grad_norm": 0.18688088655471802, + "learning_rate": 1.1327335742050442e-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": 0.005779878236353397, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.002441789919976145, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01400154661387205, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.01671588756144047, + "weighted_rkd_loss": 0.007000773306936025 + } + ], + "logging_steps": 10, + "max_steps": 5000, + "num_input_tokens_seen": 0, + "num_train_epochs": 9, + "save_steps": 5001, + "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": 64, + "trial_name": null, + "trial_params": null +} diff --git a/FewShot/39tasks/5shot/FromRKD-Timewar_Processingline_v2/seed0/default/train_5shot_seed0_processingline_5k/checkpoint-5000/config.json b/FewShot/39tasks/5shot/FromRKD-Timewar_Processingline_v2/seed0/default/train_5shot_seed0_processingline_5k/checkpoint-5000/config.json new file mode 100644 index 0000000000000000000000000000000000000000..4740d0455bef31c469ee5dab6d4a47d64df61665 --- /dev/null +++ b/FewShot/39tasks/5shot/FromRKD-Timewar_Processingline_v2/seed0/default/train_5shot_seed0_processingline_5k/checkpoint-5000/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": 0.5, + "rkd_loss_weight_end": 0.0, + "rkd_loss_weight_schedule": null, + "rkd_loss_weight_start": 0.1, + "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/FewShot/39tasks/5shot/FromRKD-Timewar_Processingline_v2/seed0/default/train_5shot_seed0_processingline_5k/checkpoint-5000/experiment_cfg/metadata.json b/FewShot/39tasks/5shot/FromRKD-Timewar_Processingline_v2/seed0/default/train_5shot_seed0_processingline_5k/checkpoint-5000/experiment_cfg/metadata.json new file mode 100644 index 0000000000000000000000000000000000000000..6f73f8489294c58acc82d1351a17e5e3ff316e18 --- /dev/null +++ b/FewShot/39tasks/5shot/FromRKD-Timewar_Processingline_v2/seed0/default/train_5shot_seed0_processingline_5k/checkpoint-5000/experiment_cfg/metadata.json @@ -0,0 +1,431 @@ +{ + "new_embodiment": { + "statistics": { + "state": { + "base_position": { + "max": [ + 7.778976917266846, + 0.33403685688972473, + 0.7223060727119446 + ], + "min": [ + -4.821934223175049, + -6.785110950469971, + 0.6994727253913879 + ], + "mean": [ + 2.5886094736594543, + -2.1559664577316586, + 0.7003580854548973 + ], + "std": [ + 1.6717149343420803, + 1.3743695399555862, + 0.000940069542359506 + ], + "q01": [ + -0.7517815060249613, + -5.497849509089705, + 0.6999776914424374 + ], + "q99": [ + 6.060945690807028, + -0.5331499932493624, + 0.7028864260472257 + ] + }, + "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, + 0.24488939171811902, + 0.5968549204077488 + ], + "std": [ + 0.0, + 0.0, + 0.6659970734507764, + 0.37443298392262264 + ], + "q01": [ + 0.0, + 0.0, + -0.9999999999999999, + 4.149924358336069e-06 + ], + "q99": [ + 0.0, + 0.0, + 0.9999999999999999, + 0.9999999999999999 + ] + }, + "end_effector_position_relative": { + "max": [ + 0.8848381638526917, + 0.8569459915161133, + 1.053884744644165 + ], + "min": [ + -0.5828549861907959, + -0.8486236333847046, + -0.3446122705936432 + ], + "mean": [ + 0.25087486209693133, + -0.005738920462036393, + 0.43643948866361787 + ], + "std": [ + 0.15741544532518012, + 0.2547603106793105, + 0.2050434492495558 + ], + "q01": [ + -0.013990057288082561, + -0.4824599092623036, + 0.18825220353500022 + ], + "q99": [ + 0.5411989642841469, + 0.4754923657245658, + 0.6912323747882034 + ] + }, + "end_effector_rotation_relative": { + "max": [ + 0.9999949932098389, + 0.9997354745864868, + 0.9900747537612915, + 0.9735639095306396 + ], + "min": [ + -0.9999914169311523, + -0.9999884366989136, + -0.9975369572639465, + 6.754255821306288e-08 + ], + "mean": [ + -0.22290331318740714, + 0.04579037037503981, + -0.05804094906527823, + 0.12347219024917105 + ], + "std": [ + 0.8519267931761344, + 0.2949394732270141, + 0.30026078832556913, + 0.16333086331829588 + ], + "q01": [ + -0.9957959450418518, + -0.5678869493126506, + -0.5318135475761875, + 0.0014160465686576505 + ], + "q99": [ + 0.9900692296898317, + 0.6357358195673684, + 0.5009847868015971, + 0.4287044482210845 + ] + }, + "gripper_qpos": { + "max": [ + 0.05289580300450325, + 0.00937709677964449 + ], + "min": [ + -0.011669117026031017, + -0.06307756155729294 + ], + "mean": [ + 0.03203268930378034, + -0.032331639271968075 + ], + "std": [ + 0.011607049173153697, + 0.011306286745872529 + ], + "q01": [ + 0.006429109393656672, + -0.04056586943958638 + ], + "q99": [ + 0.040377860142451545, + -0.006593889698673258 + ] + } + }, + "action": { + "base_motion": { + "max": [ + 1.0, + 1.0, + 1.0, + 0.0 + ], + "min": [ + -1.0, + -1.0, + -1.0, + 0.0 + ], + "mean": [ + 0.006050336852760097, + 0.002728916466628117, + -0.0017236811393598838, + 0.0 + ], + "std": [ + 0.11032167711449951, + 0.09824140301615687, + 0.07825208047515658, + 0.0 + ], + "q01": [ + -0.154721262094005, + -0.18955302988697728, + -0.14454065867754942, + 0.0 + ], + "q99": [ + 0.24492955247744197, + 0.2496797822190429, + 0.11966742325336034, + 0.0 + ] + }, + "control_mode": { + "max": [ + 1.0 + ], + "min": [ + -1.0 + ], + "mean": [ + -0.9294448715752568 + ], + "std": [ + 0.36897903503358326 + ], + "q01": [ + -1.0 + ], + "q99": [ + -0.21627562513311385 + ] + }, + "end_effector_position": { + "max": [ + 1.0, + 1.0, + 1.0 + ], + "min": [ + -1.0, + -1.0, + -1.0 + ], + "mean": [ + -0.00826780165926883, + 0.003161997359379232, + -0.08157453651841007 + ], + "std": [ + 0.4545430944464769, + 0.41040151434986594, + 0.4102644251536544 + ], + "q01": [ + -0.9653185255999107, + -0.9255635199318754, + -0.8208437313341718 + ], + "q99": [ + 0.7987781084592441, + 0.9260844232330088, + 0.8357421812622201 + ] + }, + "end_effector_rotation": { + "max": [ + 1.0, + 1.0, + 1.0 + ], + "min": [ + -1.0, + -1.0, + -1.0 + ], + "mean": [ + 0.007367024641295137, + -0.02060756343728361, + -0.001884833075998489 + ], + "std": [ + 0.1116768956941076, + 0.13249042272165212, + 0.12338908364725322 + ], + "q01": [ + -0.26894658663246845, + -0.39865607740339326, + -0.34393735191302655 + ], + "q99": [ + 0.3191674780171587, + 0.31133630397431267, + 0.33968476423051097 + ] + }, + "gripper_close": { + "max": [ + 1.0 + ], + "min": [ + -1.0 + ], + "mean": [ + -0.3098994894722959 + ], + "std": [ + 0.9507718287498876 + ], + "q01": [ + -1.0 + ], + "q99": [ + 0.7545620899936146 + ] + } + } + }, + "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/FewShot/39tasks/5shot/FromRKD-Timewar_Processingline_v2/seed0/default/train_5shot_seed0_processingline_5k/checkpoint-5000/model.safetensors.index.json b/FewShot/39tasks/5shot/FromRKD-Timewar_Processingline_v2/seed0/default/train_5shot_seed0_processingline_5k/checkpoint-5000/model.safetensors.index.json new file mode 100644 index 0000000000000000000000000000000000000000..b20baed71f4d16a411223b70d45d934c6cbbfe97 --- /dev/null +++ b/FewShot/39tasks/5shot/FromRKD-Timewar_Processingline_v2/seed0/default/train_5shot_seed0_processingline_5k/checkpoint-5000/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/FewShot/39tasks/5shot/FromRKD-Timewar_Processingline_v2/seed0/default/train_5shot_seed0_processingline_5k/checkpoint-5000/trainer_state.json b/FewShot/39tasks/5shot/FromRKD-Timewar_Processingline_v2/seed0/default/train_5shot_seed0_processingline_5k/checkpoint-5000/trainer_state.json new file mode 100644 index 0000000000000000000000000000000000000000..e7bc9d67ef6fb6747218daeb7a5b7d2b7e4d535e --- /dev/null +++ b/FewShot/39tasks/5shot/FromRKD-Timewar_Processingline_v2/seed0/default/train_5shot_seed0_processingline_5k/checkpoint-5000/trainer_state.json @@ -0,0 +1,23034 @@ +{ + "best_global_step": null, + "best_metric": null, + "best_model_checkpoint": null, + "epoch": 8.389261744966444, + "eval_steps": 500, + "global_step": 5000, + "is_hyper_param_search": false, + "is_local_process_zero": true, + "is_world_process_zero": true, + "log_history": [ + { + "epoch": 0.016778523489932886, + "flow_matching_loss": 0.2056944653391838, + "grad_norm": 0.9650156497955322, + "learning_rate": 1.08e-06, + "loss": 0.3957, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.13596224263310433, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.10807111635804176, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.37999560236930846, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.2056944653391838, + "weighted_rkd_loss": 0.18999780118465423 + }, + { + "epoch": 0.03355704697986577, + "flow_matching_loss": 0.24405399858951568, + "grad_norm": 1.1819026470184326, + "learning_rate": 2.28e-06, + "loss": 0.4373, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.13781716972589492, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.11078641712665557, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.3864207535982132, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.24405399858951568, + "weighted_rkd_loss": 0.1932103767991066 + }, + { + "epoch": 0.050335570469798654, + "flow_matching_loss": 0.18607925474643708, + "grad_norm": 1.1378169059753418, + "learning_rate": 3.48e-06, + "loss": 0.3642, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.12805165946483613, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.10023756325244904, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.3563408851623535, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.18607925474643708, + "weighted_rkd_loss": 0.17817044258117676 + }, + { + "epoch": 0.06711409395973154, + "flow_matching_loss": 0.1954496458172798, + "grad_norm": 0.9604244232177734, + "learning_rate": 4.68e-06, + "loss": 0.3626, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.12190236821770668, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.09053741469979286, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.33434214890003205, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.1954496458172798, + "weighted_rkd_loss": 0.16717107445001603 + }, + { + "epoch": 0.08389261744966443, + "flow_matching_loss": 0.18862971812486648, + "grad_norm": 1.490982174873352, + "learning_rate": 5.8800000000000005e-06, + "loss": 0.3556, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.12135515362024307, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.09131846278905868, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.33402876555919647, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.18862971812486648, + "weighted_rkd_loss": 0.16701438277959824 + }, + { + "epoch": 0.10067114093959731, + "flow_matching_loss": 0.1686223641037941, + "grad_norm": 0.6995788812637329, + "learning_rate": 7.08e-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": 0.11892070844769478, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.08481515794992447, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.322656574845314, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.1686223641037941, + "weighted_rkd_loss": 0.161328287422657 + }, + { + "epoch": 0.1174496644295302, + "flow_matching_loss": 0.1650075815618038, + "grad_norm": 0.7736762762069702, + "learning_rate": 8.28e-06, + "loss": 0.3138, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.10964272096753121, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.07826093062758446, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.2975463718175888, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.1650075815618038, + "weighted_rkd_loss": 0.1487731859087944 + }, + { + "epoch": 0.1342281879194631, + "flow_matching_loss": 0.14050063565373422, + "grad_norm": 0.7896946668624878, + "learning_rate": 9.48e-06, + "loss": 0.2771, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.10285875871777535, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.06756566911935806, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.27328317910432814, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.14050063565373422, + "weighted_rkd_loss": 0.13664158955216407 + }, + { + "epoch": 0.15100671140939598, + "flow_matching_loss": 0.12970626056194307, + "grad_norm": 0.6709607839584351, + "learning_rate": 1.068e-05, + "loss": 0.2633, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.09869295060634613, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.06970738619565964, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.267093288898468, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.12970626056194307, + "weighted_rkd_loss": 0.133546644449234 + }, + { + "epoch": 0.16778523489932887, + "flow_matching_loss": 0.11140652000904083, + "grad_norm": 0.6956789493560791, + "learning_rate": 1.1880000000000001e-05, + "loss": 0.2261, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.0861569106578827, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.05709078311920166, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.22940460294485093, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.11140652000904083, + "weighted_rkd_loss": 0.11470230147242547 + }, + { + "epoch": 0.18456375838926176, + "flow_matching_loss": 0.09781593903899193, + "grad_norm": 0.7636014223098755, + "learning_rate": 1.308e-05, + "loss": 0.2177, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.09013360589742661, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.059417485073208806, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.23968469649553298, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.09781593903899193, + "weighted_rkd_loss": 0.11984234824776649 + }, + { + "epoch": 0.20134228187919462, + "flow_matching_loss": 0.08636336326599121, + "grad_norm": 0.6833075881004333, + "learning_rate": 1.428e-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": 0.0791482873260975, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.052746016159653665, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.2110425904393196, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.08636336326599121, + "weighted_rkd_loss": 0.1055212952196598 + }, + { + "epoch": 0.2181208053691275, + "flow_matching_loss": 0.08181707337498664, + "grad_norm": 0.5722662210464478, + "learning_rate": 1.548e-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": 0.0808029130101204, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.05019080862402916, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.2117966368794441, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.08181707337498664, + "weighted_rkd_loss": 0.10589831843972206 + }, + { + "epoch": 0.2348993288590604, + "flow_matching_loss": 0.08008071929216384, + "grad_norm": 0.5205653309822083, + "learning_rate": 1.6680000000000003e-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": 0.07563451901078225, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.046241700649261475, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.19751073718070983, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.08008071929216384, + "weighted_rkd_loss": 0.09875536859035491 + }, + { + "epoch": 0.2516778523489933, + "flow_matching_loss": 0.07708811983466149, + "grad_norm": 0.6222124099731445, + "learning_rate": 1.7879999999999998e-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": 0.07161490805447102, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.04470515791326761, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.18793497681617738, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.07708811983466149, + "weighted_rkd_loss": 0.09396748840808869 + }, + { + "epoch": 0.2684563758389262, + "flow_matching_loss": 0.07064299248158931, + "grad_norm": 0.5130728483200073, + "learning_rate": 1.908e-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": 0.06377769745886326, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.038518281653523445, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.166073676943779, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.07064299248158931, + "weighted_rkd_loss": 0.0830368384718895 + }, + { + "epoch": 0.28523489932885904, + "flow_matching_loss": 0.0708860356360674, + "grad_norm": 0.44459694623947144, + "learning_rate": 2.0280000000000002e-05, + "loss": 0.1507, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.06139358542859554, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.03676006477326155, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.15954723656177522, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.0708860356360674, + "weighted_rkd_loss": 0.07977361828088761 + }, + { + "epoch": 0.30201342281879195, + "flow_matching_loss": 0.06621370911598205, + "grad_norm": 0.43706104159355164, + "learning_rate": 2.148e-05, + "loss": 0.1474, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.062162948027253154, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.03814310748130083, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.16246900260448455, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.06621370911598205, + "weighted_rkd_loss": 0.08123450130224227 + }, + { + "epoch": 0.3187919463087248, + "flow_matching_loss": 0.06586643978953362, + "grad_norm": 0.48638027906417847, + "learning_rate": 2.268e-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": 0.058573270216584206, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.03490185085684061, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.15204839110374452, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.06586643978953362, + "weighted_rkd_loss": 0.07602419555187226 + }, + { + "epoch": 0.33557046979865773, + "flow_matching_loss": 0.06912102214992047, + "grad_norm": 0.45584234595298767, + "learning_rate": 2.3880000000000002e-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": 0.0558802418410778, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.03365606628358364, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.1454165495932102, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.06912102214992047, + "weighted_rkd_loss": 0.0727082747966051 + }, + { + "epoch": 0.3523489932885906, + "flow_matching_loss": 0.06224445998668671, + "grad_norm": 0.516363799571991, + "learning_rate": 2.508e-05, + "loss": 0.1261, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.04999670423567295, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.027720438688993453, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.12771384567022323, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.06224445998668671, + "weighted_rkd_loss": 0.06385692283511162 + }, + { + "epoch": 0.3691275167785235, + "flow_matching_loss": 0.06367993578314782, + "grad_norm": 0.4426209032535553, + "learning_rate": 2.628e-05, + "loss": 0.1292, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.050844109430909154, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.02938362695276737, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.1310718484222889, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.06367993578314782, + "weighted_rkd_loss": 0.06553592421114444 + }, + { + "epoch": 0.3859060402684564, + "flow_matching_loss": 0.061497534066438674, + "grad_norm": 0.4488156735897064, + "learning_rate": 2.748e-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": 0.04617831408977509, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.02737251725047827, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.11972914338111877, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.061497534066438674, + "weighted_rkd_loss": 0.059864571690559386 + }, + { + "epoch": 0.40268456375838924, + "flow_matching_loss": 0.05670501254498959, + "grad_norm": 0.4305253028869629, + "learning_rate": 2.868e-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": 0.04511659294366836, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.02554646022617817, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.11577964574098587, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.05670501254498959, + "weighted_rkd_loss": 0.057889822870492935 + }, + { + "epoch": 0.41946308724832215, + "flow_matching_loss": 0.052361448854207994, + "grad_norm": 0.39369747042655945, + "learning_rate": 2.9880000000000002e-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": 0.04291637688875198, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.021558509021997452, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.10739126279950142, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.052361448854207994, + "weighted_rkd_loss": 0.05369563139975071 + }, + { + "epoch": 0.436241610738255, + "flow_matching_loss": 0.05451476089656353, + "grad_norm": 0.5052051544189453, + "learning_rate": 2.9999734259635676e-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": 0.04195186048746109, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.02312448900192976, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.10702821090817452, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.05451476089656353, + "weighted_rkd_loss": 0.05351410545408726 + }, + { + "epoch": 0.45302013422818793, + "flow_matching_loss": 0.060980511084198954, + "grad_norm": 0.4307898283004761, + "learning_rate": 2.9998815663057245e-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": 0.04224815517663956, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.023023180663585663, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.10751949027180671, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.060980511084198954, + "weighted_rkd_loss": 0.053759745135903356 + }, + { + "epoch": 0.4697986577181208, + "flow_matching_loss": 0.050894477590918544, + "grad_norm": 0.395856112241745, + "learning_rate": 2.999724096969228e-05, + "loss": 0.1016, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.039661315828561784, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.022144765593111515, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.10146739631891251, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.050894477590918544, + "weighted_rkd_loss": 0.050733698159456254 + }, + { + "epoch": 0.4865771812080537, + "flow_matching_loss": 0.05287523865699768, + "grad_norm": 0.3969660699367523, + "learning_rate": 2.9995010248422984e-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": 0.03771056812256575, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.019884974136948584, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.09530610963702202, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.05287523865699768, + "weighted_rkd_loss": 0.04765305481851101 + }, + { + "epoch": 0.5033557046979866, + "flow_matching_loss": 0.0479743555188179, + "grad_norm": 0.36045897006988525, + "learning_rate": 2.999212359682832e-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": 0.03238320779055357, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.016917126439511775, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.08168354257941246, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.0479743555188179, + "weighted_rkd_loss": 0.04084177128970623 + }, + { + "epoch": 0.5201342281879194, + "flow_matching_loss": 0.04859849065542221, + "grad_norm": 0.39492926001548767, + "learning_rate": 2.9988581141179787e-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": 0.034291915595531464, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.016848549619317053, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.08543238118290901, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.04859849065542221, + "weighted_rkd_loss": 0.042716190591454505 + }, + { + "epoch": 0.5369127516778524, + "flow_matching_loss": 0.04939671419560909, + "grad_norm": 0.4409841299057007, + "learning_rate": 2.998438303643587e-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": 0.03414266463369131, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.018428620789200066, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.08671394959092141, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.04939671419560909, + "weighted_rkd_loss": 0.043356974795460704 + }, + { + "epoch": 0.5536912751677853, + "flow_matching_loss": 0.04955456294119358, + "grad_norm": 0.37391889095306396, + "learning_rate": 2.997952946623527e-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": 0.033923293277621266, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.018506798055022956, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.0863533854484558, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.04955456294119358, + "weighted_rkd_loss": 0.0431766927242279 + }, + { + "epoch": 0.5704697986577181, + "flow_matching_loss": 0.050061154365539554, + "grad_norm": 0.40724676847457886, + "learning_rate": 2.9974020642888875e-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": 0.03251873888075352, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.018702817521989344, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.08374029472470283, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.050061154365539554, + "weighted_rkd_loss": 0.041870147362351416 + }, + { + "epoch": 0.587248322147651, + "flow_matching_loss": 0.04783603325486183, + "grad_norm": 0.3322409391403198, + "learning_rate": 2.9967856807370458e-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": 0.02946928795427084, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.016537870932370424, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.07547644712030888, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.04783603325486183, + "weighted_rkd_loss": 0.03773822356015444 + }, + { + "epoch": 0.6040268456375839, + "flow_matching_loss": 0.04893294684588909, + "grad_norm": 0.37779635190963745, + "learning_rate": 2.996103822930615e-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": 0.03286735378205776, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.01721991701051593, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.08295462429523467, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.04893294684588909, + "weighted_rkd_loss": 0.04147731214761734 + }, + { + "epoch": 0.6208053691275168, + "flow_matching_loss": 0.04373093023896217, + "grad_norm": 0.3719472587108612, + "learning_rate": 2.9953565206962653e-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": 0.030151859484612943, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.015161976683884859, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.07546569555997848, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.04373093023896217, + "weighted_rkd_loss": 0.03773284777998924 + }, + { + "epoch": 0.6375838926174496, + "flow_matching_loss": 0.0451244730502367, + "grad_norm": 0.41764843463897705, + "learning_rate": 2.9945438067234172e-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": 0.031219374760985373, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.016777731571346523, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.07921648174524307, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.0451244730502367, + "weighted_rkd_loss": 0.039608240872621533 + }, + { + "epoch": 0.6543624161073825, + "flow_matching_loss": 0.04234543144702911, + "grad_norm": 0.3251534402370453, + "learning_rate": 2.9936657165628124e-05, + "loss": 0.0805, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.03002563174813986, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.01633982127532363, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.07639108411967754, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.04234543144702911, + "weighted_rkd_loss": 0.03819554205983877 + }, + { + "epoch": 0.6711409395973155, + "flow_matching_loss": 0.04180084690451622, + "grad_norm": 0.3378938138484955, + "learning_rate": 2.992722288624959e-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": 0.026612223125994205, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.014572171401232481, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.06779661700129509, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.04180084690451622, + "weighted_rkd_loss": 0.033898308500647546 + }, + { + "epoch": 0.6879194630872483, + "flow_matching_loss": 0.04224371202290058, + "grad_norm": 0.3396628797054291, + "learning_rate": 2.9917135641784515e-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": 0.025039970688521863, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0127363421022892, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.06281628347933292, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.04224371202290058, + "weighted_rkd_loss": 0.03140814173966646 + }, + { + "epoch": 0.7046979865771812, + "flow_matching_loss": 0.03971978519111872, + "grad_norm": 0.30670905113220215, + "learning_rate": 2.9906395873481655e-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": 0.02627517990767956, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.013494232017546893, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.06604459211230278, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.03971978519111872, + "weighted_rkd_loss": 0.03302229605615139 + }, + { + "epoch": 0.7214765100671141, + "flow_matching_loss": 0.04066108614206314, + "grad_norm": 0.39300358295440674, + "learning_rate": 2.989500405113326e-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": 0.026304544880986214, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.01363995224237442, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.06624904237687587, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.04066108614206314, + "weighted_rkd_loss": 0.033124521188437936 + }, + { + "epoch": 0.738255033557047, + "flow_matching_loss": 0.04158542156219482, + "grad_norm": 0.3305646777153015, + "learning_rate": 2.9882960673054547e-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": 0.025151407718658446, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.01327982721850276, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.06358264274895191, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.04158542156219482, + "weighted_rkd_loss": 0.031791321374475955 + }, + { + "epoch": 0.7550335570469798, + "flow_matching_loss": 0.04144757613539696, + "grad_norm": 0.34595179557800293, + "learning_rate": 2.9870266266061886e-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": 0.022307384945452213, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0116576106287539, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.0562723807990551, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.04144757613539696, + "weighted_rkd_loss": 0.02813619039952755 + }, + { + "epoch": 0.7718120805369127, + "flow_matching_loss": 0.04313542582094669, + "grad_norm": 0.3158428370952606, + "learning_rate": 2.9856921385449766e-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": 0.02726253420114517, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.014133189246058464, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.0686582576483488, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.04313542582094669, + "weighted_rkd_loss": 0.0343291288241744 + }, + { + "epoch": 0.7885906040268457, + "flow_matching_loss": 0.042977604642510416, + "grad_norm": 0.3631656765937805, + "learning_rate": 2.9842926614966502e-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": 0.023044148832559584, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.012475840467959642, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.05856413841247558, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.042977604642510416, + "weighted_rkd_loss": 0.02928206920623779 + }, + { + "epoch": 0.8053691275167785, + "flow_matching_loss": 0.03909936882555485, + "grad_norm": 0.29421156644821167, + "learning_rate": 2.982828256678868e-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": 0.02391676288098097, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0120180937461555, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.059851620346307755, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.03909936882555485, + "weighted_rkd_loss": 0.029925810173153877 + }, + { + "epoch": 0.8221476510067114, + "flow_matching_loss": 0.03855758477002382, + "grad_norm": 0.2879815399646759, + "learning_rate": 2.9812989881494412e-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": 0.02446096483618021, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.012417732179164887, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.061339662224054334, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.03855758477002382, + "weighted_rkd_loss": 0.030669831112027167 + }, + { + "epoch": 0.8389261744966443, + "flow_matching_loss": 0.04289716742932796, + "grad_norm": 0.37124210596084595, + "learning_rate": 2.9797049228035296e-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": 0.0244005236774683, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.013081625755876302, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.06188267394900322, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.04289716742932796, + "weighted_rkd_loss": 0.03094133697450161 + }, + { + "epoch": 0.8557046979865772, + "flow_matching_loss": 0.03983708620071411, + "grad_norm": 0.3242025375366211, + "learning_rate": 2.978046130370715e-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": 0.022115133143961428, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.011383708892390132, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.05561397559940815, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.03983708620071411, + "weighted_rkd_loss": 0.027806987799704076 + }, + { + "epoch": 0.87248322147651, + "flow_matching_loss": 0.03861053548753261, + "grad_norm": 0.30338215827941895, + "learning_rate": 2.9763226834119535e-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": 0.021896894834935666, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.011009396938607096, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.054803187027573586, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.03861053548753261, + "weighted_rkd_loss": 0.027401593513786793 + }, + { + "epoch": 0.889261744966443, + "flow_matching_loss": 0.037504769116640094, + "grad_norm": 0.2798364460468292, + "learning_rate": 2.9745346573163963e-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": 0.019946480728685857, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.009748974908143282, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.04964193664491177, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.037504769116640094, + "weighted_rkd_loss": 0.024820968322455884 + }, + { + "epoch": 0.9060402684563759, + "flow_matching_loss": 0.03797667548060417, + "grad_norm": 0.25551819801330566, + "learning_rate": 2.9726821302980987e-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": 0.021121531259268522, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.01155223953537643, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.05379530191421509, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.03797667548060417, + "weighted_rkd_loss": 0.026897650957107545 + }, + { + "epoch": 0.9228187919463087, + "flow_matching_loss": 0.037000850029289725, + "grad_norm": 0.31982243061065674, + "learning_rate": 2.9707651833925935e-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": 0.0204544547945261, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.009314438654109835, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.05022334791719914, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.037000850029289725, + "weighted_rkd_loss": 0.02511167395859957 + }, + { + "epoch": 0.9395973154362416, + "flow_matching_loss": 0.038409977406263354, + "grad_norm": 0.32383039593696594, + "learning_rate": 2.968783900453348e-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": 0.01836938913911581, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.010035208612680434, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.046773987263441084, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.038409977406263354, + "weighted_rkd_loss": 0.023386993631720542 + }, + { + "epoch": 0.9563758389261745, + "flow_matching_loss": 0.03884415850043297, + "grad_norm": 0.3668201267719269, + "learning_rate": 2.9667383681480968e-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": 0.01754778316244483, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.008952949335798622, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.04404851570725441, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.03884415850043297, + "weighted_rkd_loss": 0.022024257853627206 + }, + { + "epoch": 0.9731543624161074, + "flow_matching_loss": 0.03858929816633463, + "grad_norm": 0.31969910860061646, + "learning_rate": 2.9646286759550493e-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": 0.018698493484407663, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.00886254720389843, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.04625953398644924, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.03858929816633463, + "weighted_rkd_loss": 0.02312976699322462 + }, + { + "epoch": 0.9899328859060402, + "flow_matching_loss": 0.037984182313084605, + "grad_norm": 0.37169548869132996, + "learning_rate": 2.9624549161589764e-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": 0.020831290166825056, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.011616774275898934, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.0532793540507555, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.037984182313084605, + "weighted_rkd_loss": 0.02663967702537775 + }, + { + "epoch": 1.0067114093959733, + "flow_matching_loss": 0.032401621155440805, + "grad_norm": 0.2810411751270294, + "learning_rate": 2.9602171838471745e-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": 0.02296459171921015, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.010374387260526419, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.05630357041954994, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.032401621155440805, + "weighted_rkd_loss": 0.02815178520977497 + }, + { + "epoch": 1.023489932885906, + "flow_matching_loss": 0.03286149390041828, + "grad_norm": 0.2521582245826721, + "learning_rate": 2.957915576905303e-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": 0.017456885334104298, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.008559525944292545, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.04347329661250114, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.03286149390041828, + "weighted_rkd_loss": 0.02173664830625057 + }, + { + "epoch": 1.0402684563758389, + "flow_matching_loss": 0.03205829132348299, + "grad_norm": 0.3185127377510071, + "learning_rate": 2.9555501960131058e-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": 0.01915504578500986, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.009150991961359978, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.04746108390390873, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.03205829132348299, + "weighted_rkd_loss": 0.023730541951954363 + }, + { + "epoch": 1.0570469798657718, + "flow_matching_loss": 0.03440665490925312, + "grad_norm": 0.2903192341327667, + "learning_rate": 2.9531211446400068e-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": 0.018206587247550487, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.009964327700436116, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.04637750200927258, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.03440665490925312, + "weighted_rkd_loss": 0.02318875100463629 + }, + { + "epoch": 1.0738255033557047, + "flow_matching_loss": 0.0327774690464139, + "grad_norm": 0.2796550691127777, + "learning_rate": 2.9506285290405813e-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": 0.01788708958774805, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.009063597302883863, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.044837776198983194, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.0327774690464139, + "weighted_rkd_loss": 0.022418888099491597 + }, + { + "epoch": 1.0906040268456376, + "flow_matching_loss": 0.03253862913697958, + "grad_norm": 0.29687777161598206, + "learning_rate": 2.948072458249911e-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": 0.018407645542174576, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.008646867936477065, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.04546215943992138, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.03253862913697958, + "weighted_rkd_loss": 0.02273107971996069 + }, + { + "epoch": 1.1073825503355705, + "flow_matching_loss": 0.034775618463754654, + "grad_norm": 0.26023975014686584, + "learning_rate": 2.9454530440788125e-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": 0.017261851113289595, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.008692631544545294, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.04321633391082287, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.034775618463754654, + "weighted_rkd_loss": 0.021608166955411436 + }, + { + "epoch": 1.1241610738255035, + "flow_matching_loss": 0.03697704430669546, + "grad_norm": 0.2973390817642212, + "learning_rate": 2.9427704011089474e-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": 0.017315912712365388, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.008650409011170267, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.043282233737409116, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.03697704430669546, + "weighted_rkd_loss": 0.021641116868704558 + }, + { + "epoch": 1.1409395973154361, + "flow_matching_loss": 0.034938321262598035, + "grad_norm": 0.33880263566970825, + "learning_rate": 2.94002464668781e-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": 0.01774648055434227, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.008731141500175, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.04422410242259502, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.034938321262598035, + "weighted_rkd_loss": 0.02211205121129751 + }, + { + "epoch": 1.157718120805369, + "flow_matching_loss": 0.03397301379591226, + "grad_norm": 0.30417579412460327, + "learning_rate": 2.9372159009235936e-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": 0.016883174236863852, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0084213403519243, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.04218768924474716, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.03397301379591226, + "weighted_rkd_loss": 0.02109384462237358 + }, + { + "epoch": 1.174496644295302, + "flow_matching_loss": 0.033245801366865634, + "grad_norm": 0.2966938614845276, + "learning_rate": 2.9343442866799365e-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": 0.01730244169011712, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.008166278852149845, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.04277116190642118, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.033245801366865634, + "weighted_rkd_loss": 0.02138558095321059 + }, + { + "epoch": 1.191275167785235, + "flow_matching_loss": 0.032241361774504186, + "grad_norm": 0.3242097496986389, + "learning_rate": 2.9314099295705495e-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": 0.01751971701160073, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.008714311430230737, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.04375374540686607, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.032241361774504186, + "weighted_rkd_loss": 0.021876872703433036 + }, + { + "epoch": 1.2080536912751678, + "flow_matching_loss": 0.030840910412371157, + "grad_norm": 0.3416939675807953, + "learning_rate": 2.928412957953718e-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": 0.016577553190290927, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.008127822168171405, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.041282928362488744, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.030840910412371157, + "weighted_rkd_loss": 0.020641464181244372 + }, + { + "epoch": 1.2248322147651007, + "flow_matching_loss": 0.030589586496353148, + "grad_norm": 0.3434118926525116, + "learning_rate": 2.925353502926689e-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": 0.014619570132344962, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.00703063509427011, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.03626977521926165, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.030589586496353148, + "weighted_rkd_loss": 0.018134887609630824 + }, + { + "epoch": 1.2416107382550337, + "flow_matching_loss": 0.028109255991876126, + "grad_norm": 0.2770143449306488, + "learning_rate": 2.922231698319937e-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": 0.015383913926780225, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.007206043181940913, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.03797387108206749, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.028109255991876126, + "weighted_rkd_loss": 0.018986935541033746 + }, + { + "epoch": 1.2583892617449663, + "flow_matching_loss": 0.03261192515492439, + "grad_norm": 0.271056592464447, + "learning_rate": 2.9190476806913088e-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": 0.016976217925548553, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.007505776965990663, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.04145821258425712, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.03261192515492439, + "weighted_rkd_loss": 0.02072910629212856 + }, + { + "epoch": 1.2751677852348993, + "flow_matching_loss": 0.030070590786635876, + "grad_norm": 0.2317705899477005, + "learning_rate": 2.9158015893200504e-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": 0.016096030082553624, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.007664719317108393, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.03985677920281887, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.030070590786635876, + "weighted_rkd_loss": 0.019928389601409434 + }, + { + "epoch": 1.2919463087248322, + "flow_matching_loss": 0.033119112625718115, + "grad_norm": 0.2506316006183624, + "learning_rate": 2.9124935662007137e-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": 0.01613023495301604, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.007737922063097358, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.03999839182943106, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.033119112625718115, + "weighted_rkd_loss": 0.01999919591471553 + }, + { + "epoch": 1.308724832214765, + "flow_matching_loss": 0.031553388386964795, + "grad_norm": 0.2889447808265686, + "learning_rate": 2.9091237560369467e-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": 0.01787243187427521, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.009260696731507779, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.04500555992126465, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.031553388386964795, + "weighted_rkd_loss": 0.022502779960632324 + }, + { + "epoch": 1.325503355704698, + "flow_matching_loss": 0.03222518637776375, + "grad_norm": 0.2785235047340393, + "learning_rate": 2.905692306235163e-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": 0.015721357706934214, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.007582925446331501, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.039025640860199926, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.03222518637776375, + "weighted_rkd_loss": 0.019512820430099963 + }, + { + "epoch": 1.342281879194631, + "flow_matching_loss": 0.030334611982107164, + "grad_norm": 0.31459397077560425, + "learning_rate": 2.9021993668980925e-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": 0.01560999769717455, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.007159262569621206, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.03837925773113966, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.030334611982107164, + "weighted_rkd_loss": 0.01918962886556983 + }, + { + "epoch": 1.3590604026845639, + "flow_matching_loss": 0.03257175628095865, + "grad_norm": 0.2993733286857605, + "learning_rate": 2.8986450908182185e-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": 0.015799124166369437, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.00794396479614079, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.039542213454842566, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.03257175628095865, + "weighted_rkd_loss": 0.019771106727421283 + }, + { + "epoch": 1.3758389261744965, + "flow_matching_loss": 0.032540197111666204, + "grad_norm": 0.31480374932289124, + "learning_rate": 2.8950296334710902e-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": 0.01558954007923603, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.007458786852657795, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.03863786738365889, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.032540197111666204, + "weighted_rkd_loss": 0.019318933691829444 + }, + { + "epoch": 1.3926174496644295, + "flow_matching_loss": 0.03170314058661461, + "grad_norm": 0.2934443950653076, + "learning_rate": 2.8913531530085248e-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": 0.016132532991468905, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.007935636956244708, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.040200702473521234, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.03170314058661461, + "weighted_rkd_loss": 0.020100351236760617 + }, + { + "epoch": 1.4093959731543624, + "flow_matching_loss": 0.03200640454888344, + "grad_norm": 0.26509004831314087, + "learning_rate": 2.887615810251687e-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": 0.014211549051105976, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.006883570412173867, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.035306668281555174, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.03200640454888344, + "weighted_rkd_loss": 0.017653334140777587 + }, + { + "epoch": 1.4261744966442953, + "flow_matching_loss": 0.02827311120927334, + "grad_norm": 0.3237679600715637, + "learning_rate": 2.8838177686840577e-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": 0.017118581011891364, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.00820444175042212, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.04244160372763872, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.02827311120927334, + "weighted_rkd_loss": 0.02122080186381936 + }, + { + "epoch": 1.4429530201342282, + "flow_matching_loss": 0.03121668491512537, + "grad_norm": 0.3042275607585907, + "learning_rate": 2.8799591944442776e-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": 0.016424584202468395, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.007744280295446515, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.040593449398875235, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.03121668491512537, + "weighted_rkd_loss": 0.020296724699437618 + }, + { + "epoch": 1.4597315436241611, + "flow_matching_loss": 0.030370122380554677, + "grad_norm": 0.3472956120967865, + "learning_rate": 2.8760402563188848e-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": 0.017987641133368017, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.009500452037900686, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.045475734397768974, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.030370122380554677, + "weighted_rkd_loss": 0.022737867198884487 + }, + { + "epoch": 1.476510067114094, + "flow_matching_loss": 0.02861998286098242, + "grad_norm": 0.24505336582660675, + "learning_rate": 2.8720611257349285e-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": 0.0141552341170609, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.006439156364649534, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.03474962469190359, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.02861998286098242, + "weighted_rkd_loss": 0.017374812345951795 + }, + { + "epoch": 1.4932885906040267, + "flow_matching_loss": 0.03184323962777853, + "grad_norm": 0.2608340084552765, + "learning_rate": 2.868021976752471e-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": 0.01494277287274599, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.00734082362614572, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.037226369604468346, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.03184323962777853, + "weighted_rkd_loss": 0.018613184802234173 + }, + { + "epoch": 1.5100671140939599, + "flow_matching_loss": 0.03042609766125679, + "grad_norm": 0.333564430475235, + "learning_rate": 2.8639229860569738e-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": 0.0160973247140646, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.007541899429634213, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.03973654881119728, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.03042609766125679, + "weighted_rkd_loss": 0.01986827440559864 + }, + { + "epoch": 1.5268456375838926, + "flow_matching_loss": 0.03012043833732605, + "grad_norm": 0.24364671111106873, + "learning_rate": 2.8597643329515688e-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": 0.014172045607119798, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.006650603329762817, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.034994694776833056, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.03012043833732605, + "weighted_rkd_loss": 0.017497347388416528 + }, + { + "epoch": 1.5436241610738255, + "flow_matching_loss": 0.03092307262122631, + "grad_norm": 0.27232593297958374, + "learning_rate": 2.855546199349214e-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": 0.014154891762882472, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.007071852730587125, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.03538163620978594, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.03092307262122631, + "weighted_rkd_loss": 0.01769081810489297 + }, + { + "epoch": 1.5604026845637584, + "flow_matching_loss": 0.027929746173322202, + "grad_norm": 0.33788374066352844, + "learning_rate": 2.85126876976474e-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": 0.012748553790152073, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.005962743982672691, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.031459851376712325, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.027929746173322202, + "weighted_rkd_loss": 0.015729925688356163 + }, + { + "epoch": 1.5771812080536913, + "flow_matching_loss": 0.028270901925861836, + "grad_norm": 0.21897122263908386, + "learning_rate": 2.8469322313067726e-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": 0.015070199593901634, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.007116558589041233, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.0372569577768445, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.028270901925861836, + "weighted_rkd_loss": 0.01862847888842225 + }, + { + "epoch": 1.5939597315436242, + "flow_matching_loss": 0.02921714950352907, + "grad_norm": 0.3024495542049408, + "learning_rate": 2.8425367736695522e-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": 0.011751286219805479, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.005542754894122481, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.02904532738029957, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.02921714950352907, + "weighted_rkd_loss": 0.014522663690149784 + }, + { + "epoch": 1.610738255033557, + "flow_matching_loss": 0.02695902418345213, + "grad_norm": 0.2782779335975647, + "learning_rate": 2.838082589124636e-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": 0.015185461472719907, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.007375222817063332, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.03774614538997412, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.02695902418345213, + "weighted_rkd_loss": 0.01887307269498706 + }, + { + "epoch": 1.62751677852349, + "flow_matching_loss": 0.026085390895605087, + "grad_norm": 0.31670045852661133, + "learning_rate": 2.8335698725124843e-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": 0.014663305971771479, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.007240531407296658, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.03656714335083962, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.026085390895605087, + "weighted_rkd_loss": 0.01828357167541981 + }, + { + "epoch": 1.6442953020134228, + "flow_matching_loss": 0.02555117029696703, + "grad_norm": 0.2121639996767044, + "learning_rate": 2.8289988212339422e-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": 0.013465657457709312, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0062181116081774235, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.03314942680299282, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.02555117029696703, + "weighted_rkd_loss": 0.01657471340149641 + }, + { + "epoch": 1.6610738255033557, + "flow_matching_loss": 0.026020037569105627, + "grad_norm": 0.22167570888996124, + "learning_rate": 2.8243696352415997e-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": 0.01538208145648241, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.008201855747029186, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.038966018334031106, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.026020037569105627, + "weighted_rkd_loss": 0.019483009167015553 + }, + { + "epoch": 1.6778523489932886, + "flow_matching_loss": 0.029859650880098343, + "grad_norm": 0.2646207809448242, + "learning_rate": 2.8196825170310495e-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": 0.012622483912855387, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.005989065254107117, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.031234032846987246, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.029859650880098343, + "weighted_rkd_loss": 0.015617016423493623 + }, + { + "epoch": 1.6946308724832215, + "flow_matching_loss": 0.029160509817302227, + "grad_norm": 0.29436981678009033, + "learning_rate": 2.8149376716320256e-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": 0.014717372134327888, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.006617354042828083, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.036052098497748375, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.029160509817302227, + "weighted_rkd_loss": 0.018026049248874187 + }, + { + "epoch": 1.7114093959731544, + "flow_matching_loss": 0.02514624744653702, + "grad_norm": 0.2813986837863922, + "learning_rate": 2.8101353065994373e-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": 0.014104087930172681, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.007160045579075814, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.03536822162568569, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.02514624744653702, + "weighted_rkd_loss": 0.017684110812842845 + }, + { + "epoch": 1.7281879194630871, + "flow_matching_loss": 0.025560924783349037, + "grad_norm": 0.24623122811317444, + "learning_rate": 2.8052756320042885e-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": 0.013708307594060897, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.006159724202007055, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.03357633929699659, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.025560924783349037, + "weighted_rkd_loss": 0.016788169648498295 + }, + { + "epoch": 1.7449664429530203, + "flow_matching_loss": 0.02597265727818012, + "grad_norm": 0.24410276114940643, + "learning_rate": 2.8003588604244893e-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": 0.014219438191503286, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.006340369349345565, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.0347792461514473, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.02597265727818012, + "weighted_rkd_loss": 0.01738962307572365 + }, + { + "epoch": 1.761744966442953, + "flow_matching_loss": 0.02521944120526314, + "grad_norm": 0.24740788340568542, + "learning_rate": 2.7953852069355565e-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": 0.014595606923103332, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0072414461523294445, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.036432659812271596, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.02521944120526314, + "weighted_rkd_loss": 0.018216329906135798 + }, + { + "epoch": 1.778523489932886, + "flow_matching_loss": 0.027643476985394953, + "grad_norm": 0.3433672785758972, + "learning_rate": 2.790354889101206e-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": 0.013499390333890915, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.006195838330313563, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.03319461885839701, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.027643476985394953, + "weighted_rkd_loss": 0.016597309429198504 + }, + { + "epoch": 1.7953020134228188, + "flow_matching_loss": 0.02857300378382206, + "grad_norm": 0.24258039891719818, + "learning_rate": 2.7852681269638366e-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": 0.013103808369487524, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.006252429843880236, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.03246004665270448, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.02857300378382206, + "weighted_rkd_loss": 0.01623002332635224 + }, + { + "epoch": 1.8120805369127517, + "flow_matching_loss": 0.025308608263731002, + "grad_norm": 0.25732696056365967, + "learning_rate": 2.780125143034902e-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": 0.014380837325006723, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.006497780932113529, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.03525945525616407, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.025308608263731002, + "weighted_rkd_loss": 0.017629727628082036 + }, + { + "epoch": 1.8288590604026846, + "flow_matching_loss": 0.026384789682924747, + "grad_norm": 0.32120704650878906, + "learning_rate": 2.7749261622851812e-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": 0.01262503294274211, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.006175780994817614, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.031425846926867965, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.026384789682924747, + "weighted_rkd_loss": 0.015712923463433982 + }, + { + "epoch": 1.8456375838926173, + "flow_matching_loss": 0.027396404929459096, + "grad_norm": 0.21401037275791168, + "learning_rate": 2.7696714121349335e-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": 0.011610002536326647, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.005550801800563931, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.028770807012915612, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.027396404929459096, + "weighted_rkd_loss": 0.014385403506457806 + }, + { + "epoch": 1.8624161073825505, + "flow_matching_loss": 0.0279746038839221, + "grad_norm": 0.20054933428764343, + "learning_rate": 2.764361122443954e-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": 0.011772450804710389, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.005359048675745726, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.028903950192034244, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.0279746038839221, + "weighted_rkd_loss": 0.014451975096017122 + }, + { + "epoch": 1.8791946308724832, + "flow_matching_loss": 0.02448360715061426, + "grad_norm": 0.26301416754722595, + "learning_rate": 2.758995525501517e-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": 0.011649888660758733, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.00557361776009202, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.02887339498847723, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.02448360715061426, + "weighted_rkd_loss": 0.014436697494238615 + }, + { + "epoch": 1.895973154362416, + "flow_matching_loss": 0.027417019382119177, + "grad_norm": 0.22313368320465088, + "learning_rate": 2.753574856016215e-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": 0.012046439200639724, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.005405143462121486, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.029498021490871905, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.027417019382119177, + "weighted_rkd_loss": 0.014749010745435952 + }, + { + "epoch": 1.912751677852349, + "flow_matching_loss": 0.02374915238469839, + "grad_norm": 0.2883581817150116, + "learning_rate": 2.7480993511056913e-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": 0.013785356283187866, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0063822722062468525, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.03395298458635807, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.02374915238469839, + "weighted_rkd_loss": 0.016976492293179036 + }, + { + "epoch": 1.929530201342282, + "flow_matching_loss": 0.024885526299476622, + "grad_norm": 0.2205721139907837, + "learning_rate": 2.7425692502862684e-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": 0.012991191726177931, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.006324190460145473, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.032306574285030365, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.024885526299476622, + "weighted_rkd_loss": 0.016153287142515182 + }, + { + "epoch": 1.9463087248322148, + "flow_matching_loss": 0.027140559628605843, + "grad_norm": 0.32631516456604004, + "learning_rate": 2.7369847954624728e-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": 0.01370504880324006, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.006441503390669822, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.03385160099714994, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.027140559628605843, + "weighted_rkd_loss": 0.01692580049857497 + }, + { + "epoch": 1.9630872483221475, + "flow_matching_loss": 0.027510961145162584, + "grad_norm": 0.19017235934734344, + "learning_rate": 2.731346230916448e-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": 0.010960731562227011, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0051203057635575535, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.02704176902770996, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.027510961145162584, + "weighted_rkd_loss": 0.01352088451385498 + }, + { + "epoch": 1.9798657718120807, + "flow_matching_loss": 0.023988731019198895, + "grad_norm": 0.2913515865802765, + "learning_rate": 2.7256538032972746e-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": 0.010911648208275438, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0053501016926020386, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.0271733982488513, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.023988731019198895, + "weighted_rkd_loss": 0.01358669912442565 + }, + { + "epoch": 1.9966442953020134, + "flow_matching_loss": 0.023296417482197285, + "grad_norm": 0.2592824697494507, + "learning_rate": 2.7199077616101773e-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": 0.01110992943868041, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.004422898893244565, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.026642757840454577, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.023296417482197285, + "weighted_rkd_loss": 0.013321378920227289 + }, + { + "epoch": 2.0134228187919465, + "flow_matching_loss": 0.024605618230998515, + "grad_norm": 0.3102531135082245, + "learning_rate": 2.714108357205634e-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": 0.01258283476345241, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0052482510916888716, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.03041392080485821, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.024605618230998515, + "weighted_rkd_loss": 0.015206960402429104 + }, + { + "epoch": 2.030201342281879, + "flow_matching_loss": 0.027245542779564858, + "grad_norm": 0.22417199611663818, + "learning_rate": 2.7082558437683803e-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": 0.01143276859074831, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.005770996958017349, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.028636534698307514, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.027245542779564858, + "weighted_rkd_loss": 0.014318267349153757 + }, + { + "epoch": 2.046979865771812, + "flow_matching_loss": 0.023954916559159756, + "grad_norm": 0.20610450208187103, + "learning_rate": 2.702350477306315e-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": 0.011581839248538017, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.005343518941663206, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.028507197834551336, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.023954916559159756, + "weighted_rkd_loss": 0.014253598917275668 + }, + { + "epoch": 2.063758389261745, + "flow_matching_loss": 0.023449451476335526, + "grad_norm": 0.2560037672519684, + "learning_rate": 2.6963925161392963e-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": 0.010292424075305461, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.004632242536172271, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.02521709054708481, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.023449451476335526, + "weighted_rkd_loss": 0.012608545273542405 + }, + { + "epoch": 2.0805369127516777, + "flow_matching_loss": 0.02585367038846016, + "grad_norm": 0.23351788520812988, + "learning_rate": 2.6903822208878476e-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": 0.013421973027288914, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.005883032223209739, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.032726977951824665, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.02585367038846016, + "weighted_rkd_loss": 0.016363488975912333 + }, + { + "epoch": 2.097315436241611, + "flow_matching_loss": 0.02222144976258278, + "grad_norm": 0.19517207145690918, + "learning_rate": 2.6843198544617535e-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": 0.010299140820279717, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0048344668466597795, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.025432748906314372, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.02222144976258278, + "weighted_rkd_loss": 0.012716374453157186 + }, + { + "epoch": 2.1140939597315436, + "flow_matching_loss": 0.026821971125900745, + "grad_norm": 0.3056233823299408, + "learning_rate": 2.6782056820485612e-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": 0.013456127140671015, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.006366095319390297, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.03327834941446781, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.026821971125900745, + "weighted_rkd_loss": 0.016639174707233906 + }, + { + "epoch": 2.1308724832214767, + "flow_matching_loss": 0.024398398119956255, + "grad_norm": 0.2585718035697937, + "learning_rate": 2.672039971101979e-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": 0.011472947150468826, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.004935935512185097, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.027881830371916293, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.024398398119956255, + "weighted_rkd_loss": 0.013940915185958146 + }, + { + "epoch": 2.1476510067114094, + "flow_matching_loss": 0.024629255197942256, + "grad_norm": 0.1899724155664444, + "learning_rate": 2.665822991330178e-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": 0.010962147265672684, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.004843188123777509, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.026767482981085777, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.024629255197942256, + "weighted_rkd_loss": 0.013383741490542889 + }, + { + "epoch": 2.164429530201342, + "flow_matching_loss": 0.02464695330709219, + "grad_norm": 0.19738611578941345, + "learning_rate": 2.6595550146839934e-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": 0.011003640946000814, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0048983822576701645, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.026905664429068565, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.02464695330709219, + "weighted_rkd_loss": 0.013452832214534282 + }, + { + "epoch": 2.1812080536912752, + "flow_matching_loss": 0.024324053525924684, + "grad_norm": 0.21592102944850922, + "learning_rate": 2.6532363153450287e-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": 0.011682957503944635, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.00482977582141757, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.0281956909224391, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.024324053525924684, + "weighted_rkd_loss": 0.01409784546121955 + }, + { + "epoch": 2.197986577181208, + "flow_matching_loss": 0.02110758163034916, + "grad_norm": 0.18565689027309418, + "learning_rate": 2.6468671697136634e-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": 0.010184981161728501, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.004276543972082436, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.024646506272256376, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.02110758163034916, + "weighted_rkd_loss": 0.012323253136128188 + }, + { + "epoch": 2.214765100671141, + "flow_matching_loss": 0.024282259680330754, + "grad_norm": 0.2649759352207184, + "learning_rate": 2.640447856396961e-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": 0.012323899054899811, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.005656029563397169, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.030303827580064537, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.024282259680330754, + "weighted_rkd_loss": 0.015151913790032268 + }, + { + "epoch": 2.2315436241610738, + "flow_matching_loss": 0.02368467003107071, + "grad_norm": 0.1718354970216751, + "learning_rate": 2.6339786561964805e-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": 0.010672940826043486, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.005090259737335146, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.02643614150583744, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.02368467003107071, + "weighted_rkd_loss": 0.01321807075291872 + }, + { + "epoch": 2.248322147651007, + "flow_matching_loss": 0.02225311063230038, + "grad_norm": 0.1958433985710144, + "learning_rate": 2.6274598520959966e-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": 0.010571935120970011, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.004792783199809492, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.02593665346503258, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.02225311063230038, + "weighted_rkd_loss": 0.01296832673251629 + }, + { + "epoch": 2.2651006711409396, + "flow_matching_loss": 0.022798995673656463, + "grad_norm": 0.2145613580942154, + "learning_rate": 2.6208917292491185e-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": 0.010735450917854905, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.005248784064315259, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.026719686202704908, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.022798995673656463, + "weighted_rkd_loss": 0.013359843101352454 + }, + { + "epoch": 2.2818791946308723, + "flow_matching_loss": 0.0219331169500947, + "grad_norm": 0.18633633852005005, + "learning_rate": 2.6142745749668176e-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": 0.011048201052471996, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.005030419118702412, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.02712682094424963, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.0219331169500947, + "weighted_rkd_loss": 0.013563410472124815 + }, + { + "epoch": 2.2986577181208054, + "flow_matching_loss": 0.021648623794317246, + "grad_norm": 0.33190497756004333, + "learning_rate": 2.607608678704859e-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": 0.01214893334545195, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.005797731433995068, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.030095598101615904, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.021648623794317246, + "weighted_rkd_loss": 0.015047799050807952 + }, + { + "epoch": 2.315436241610738, + "flow_matching_loss": 0.022526911459863186, + "grad_norm": 0.23687943816184998, + "learning_rate": 2.600894332051139e-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": 0.009939734917134046, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.004744947282597422, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.024624417070299387, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.022526911459863186, + "weighted_rkd_loss": 0.012312208535149693 + }, + { + "epoch": 2.3322147651006713, + "flow_matching_loss": 0.02625675145536661, + "grad_norm": 0.2014767974615097, + "learning_rate": 2.594131828712932e-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": 0.011044638883322477, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0052457981742918495, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.02733507603406906, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.02625675145536661, + "weighted_rkd_loss": 0.01366753801703453 + }, + { + "epoch": 2.348993288590604, + "flow_matching_loss": 0.023217788338661192, + "grad_norm": 0.30442988872528076, + "learning_rate": 2.5873214645040414e-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": 0.01133463829755783, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.005828363331966102, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.028497640043497086, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.023217788338661192, + "weighted_rkd_loss": 0.014248820021748543 + }, + { + "epoch": 2.365771812080537, + "flow_matching_loss": 0.023031603731215, + "grad_norm": 0.25246021151542664, + "learning_rate": 2.5804635373318604e-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": 0.009690780472010375, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.004492142889648676, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.023873704113066198, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.023031603731215, + "weighted_rkd_loss": 0.011936852056533099 + }, + { + "epoch": 2.38255033557047, + "flow_matching_loss": 0.024585946835577487, + "grad_norm": 0.20293375849723816, + "learning_rate": 2.57355834718434e-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": 0.011088294768705964, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0044671180192381145, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.02664370760321617, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.024585946835577487, + "weighted_rkd_loss": 0.013321853801608085 + }, + { + "epoch": 2.3993288590604025, + "flow_matching_loss": 0.021987343579530715, + "grad_norm": 0.23088401556015015, + "learning_rate": 2.566606196116867e-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": 0.010794766712933779, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.004875816567800939, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.026465350203216075, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.021987343579530715, + "weighted_rkd_loss": 0.013232675101608037 + }, + { + "epoch": 2.4161073825503356, + "flow_matching_loss": 0.024937476590275766, + "grad_norm": 0.21601395308971405, + "learning_rate": 2.559607388239051e-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": 0.01238170089200139, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.005331674730405212, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.03009507693350315, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.024937476590275766, + "weighted_rkd_loss": 0.015047538466751576 + }, + { + "epoch": 2.4328859060402683, + "flow_matching_loss": 0.020688330102711917, + "grad_norm": 0.22601929306983948, + "learning_rate": 2.552562229701422e-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": 0.010842163767665624, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0046914838254451755, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.02637581154704094, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.020688330102711917, + "weighted_rkd_loss": 0.01318790577352047 + }, + { + "epoch": 2.4496644295302015, + "flow_matching_loss": 0.02394238654524088, + "grad_norm": 0.2014225870370865, + "learning_rate": 2.545471028682036e-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": 0.009968520933762192, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.004925135779194534, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.024862177856266498, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.02394238654524088, + "weighted_rkd_loss": 0.012431088928133249 + }, + { + "epoch": 2.466442953020134, + "flow_matching_loss": 0.02508413475006819, + "grad_norm": 0.28125983476638794, + "learning_rate": 2.538334095373e-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": 0.011306649027392267, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.005316834663972259, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.027930132672190668, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.02508413475006819, + "weighted_rkd_loss": 0.013965066336095334 + }, + { + "epoch": 2.4832214765100673, + "flow_matching_loss": 0.023891801387071608, + "grad_norm": 0.29100310802459717, + "learning_rate": 2.531151741966896e-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": 0.01073241732083261, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.005184279289096594, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.026649114117026328, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.023891801387071608, + "weighted_rkd_loss": 0.013324557058513164 + }, + { + "epoch": 2.5, + "flow_matching_loss": 0.02415292225778103, + "grad_norm": 0.22286556661128998, + "learning_rate": 2.5239242826431302e-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": 0.010552192945033312, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.004753443133085966, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.025857829488813877, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.02415292225778103, + "weighted_rkd_loss": 0.012928914744406939 + }, + { + "epoch": 2.5167785234899327, + "flow_matching_loss": 0.022898601554334163, + "grad_norm": 0.18709388375282288, + "learning_rate": 2.5166520335541865e-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": 0.01060750326141715, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.004809713829308749, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.02602472063153982, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.022898601554334163, + "weighted_rkd_loss": 0.01301236031576991 + }, + { + "epoch": 2.533557046979866, + "flow_matching_loss": 0.02009684108197689, + "grad_norm": 0.25758233666419983, + "learning_rate": 2.5093353128117996e-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": 0.010353196039795875, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.00480845645070076, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.025514848809689284, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.02009684108197689, + "weighted_rkd_loss": 0.012757424404844642 + }, + { + "epoch": 2.5503355704697985, + "flow_matching_loss": 0.02130047921091318, + "grad_norm": 0.16162274777889252, + "learning_rate": 2.501974440473037e-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": 0.009250646131113172, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0046441263752058145, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.023145418614149094, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.02130047921091318, + "weighted_rkd_loss": 0.011572709307074547 + }, + { + "epoch": 2.5671140939597317, + "flow_matching_loss": 0.0252719109877944, + "grad_norm": 0.21416917443275452, + "learning_rate": 2.4945697385263015e-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": 0.009564307797700167, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.004446143656969071, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.023574758879840375, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.0252719109877944, + "weighted_rkd_loss": 0.011787379439920188 + }, + { + "epoch": 2.5838926174496644, + "flow_matching_loss": 0.022078343108296393, + "grad_norm": 0.21463052928447723, + "learning_rate": 2.487121530877243e-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": 0.011745102750137449, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.005604003882035613, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.02909420933574438, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.022078343108296393, + "weighted_rkd_loss": 0.01454710466787219 + }, + { + "epoch": 2.600671140939597, + "flow_matching_loss": 0.021502092573791744, + "grad_norm": 0.23142777383327484, + "learning_rate": 2.4796301433345943e-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": 0.0092654038220644, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0041564338142052295, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.022687241435050964, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.021502092573791744, + "weighted_rkd_loss": 0.011343620717525482 + }, + { + "epoch": 2.61744966442953, + "flow_matching_loss": 0.022588416188955306, + "grad_norm": 0.2171335220336914, + "learning_rate": 2.4720959035959148e-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": 0.010664369678124785, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.004696646006777883, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.02602538522332907, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.022588416188955306, + "weighted_rkd_loss": 0.013012692611664534 + }, + { + "epoch": 2.6342281879194633, + "flow_matching_loss": 0.022199026308953762, + "grad_norm": 0.2863241732120514, + "learning_rate": 2.464519141233257e-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": 0.011123205535113811, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0050924249924719335, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.027338836155831815, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.022199026308953762, + "weighted_rkd_loss": 0.013669418077915907 + }, + { + "epoch": 2.651006711409396, + "flow_matching_loss": 0.02178021539002657, + "grad_norm": 0.23810994625091553, + "learning_rate": 2.456900187678754e-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": 0.010764320380985737, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.005336681194603443, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.026865322329103946, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.02178021539002657, + "weighted_rkd_loss": 0.013432661164551973 + }, + { + "epoch": 2.6677852348993287, + "flow_matching_loss": 0.022027206886559726, + "grad_norm": 0.17385552823543549, + "learning_rate": 2.4492393762101147e-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": 0.008514114422723651, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.004043441731482744, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.021071670670062305, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.022027206886559726, + "weighted_rkd_loss": 0.010535835335031152 + }, + { + "epoch": 2.684563758389262, + "flow_matching_loss": 0.022792240604758263, + "grad_norm": 0.20645810663700104, + "learning_rate": 2.4415370419360508e-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": 0.00922980452887714, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.004303267924115062, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.02276287693530321, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.022792240604758263, + "weighted_rkd_loss": 0.011381438467651606 + }, + { + "epoch": 2.7013422818791946, + "flow_matching_loss": 0.021514329593628645, + "grad_norm": 0.2190205603837967, + "learning_rate": 2.4337935217816157e-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": 0.009895919263362885, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.00468641291372478, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.02447825139388442, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.021514329593628645, + "weighted_rkd_loss": 0.01223912569694221 + }, + { + "epoch": 2.7181208053691277, + "flow_matching_loss": 0.024532474111765623, + "grad_norm": 0.3014610707759857, + "learning_rate": 2.4260091544734665e-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": 0.010453154426068067, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0046753128292039035, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.02558162175118923, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.024532474111765623, + "weighted_rkd_loss": 0.012790810875594616 + }, + { + "epoch": 2.7348993288590604, + "flow_matching_loss": 0.022621615324169397, + "grad_norm": 0.20125490427017212, + "learning_rate": 2.4181842805250466e-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": 0.008896962832659483, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.004234325815923512, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.022028251271694898, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.022621615324169397, + "weighted_rkd_loss": 0.011014125635847449 + }, + { + "epoch": 2.751677852348993, + "flow_matching_loss": 0.021766430512070657, + "grad_norm": 0.21964611113071442, + "learning_rate": 2.4103192422216923e-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": 0.009006364457309245, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.003936750791035592, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.021949479822069407, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.021766430512070657, + "weighted_rkd_loss": 0.010974739911034703 + }, + { + "epoch": 2.7684563758389262, + "flow_matching_loss": 0.022125181183218955, + "grad_norm": 0.18881672620773315, + "learning_rate": 2.402414383605658e-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": 0.009801744995638729, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.004149720002897084, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.023753209877759217, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.022125181183218955, + "weighted_rkd_loss": 0.011876604938879608 + }, + { + "epoch": 2.785234899328859, + "flow_matching_loss": 0.022959327884018423, + "grad_norm": 0.20707227289676666, + "learning_rate": 2.3944700504610674e-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": 0.010146900825202466, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.004097336274571717, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.024391137808561326, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.022959327884018423, + "weighted_rkd_loss": 0.012195568904280663 + }, + { + "epoch": 2.802013422818792, + "flow_matching_loss": 0.02108875922858715, + "grad_norm": 0.20151439309120178, + "learning_rate": 2.3864865902987883e-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": 0.008188423234969378, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.003603689349256456, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.019980535749346016, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.02108875922858715, + "weighted_rkd_loss": 0.009990267874673008 + }, + { + "epoch": 2.8187919463087248, + "flow_matching_loss": 0.02431239653378725, + "grad_norm": 0.17307604849338531, + "learning_rate": 2.378464352341231e-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": 0.00882768682204187, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.004448901535943151, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.02210427522659302, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.02431239653378725, + "weighted_rkd_loss": 0.01105213761329651 + }, + { + "epoch": 2.8355704697986575, + "flow_matching_loss": 0.021007327269762753, + "grad_norm": 0.1367725431919098, + "learning_rate": 2.3704036875070728e-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": 0.008387606404721738, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0035082216607406737, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.020283434353768826, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.021007327269762753, + "weighted_rkd_loss": 0.010141717176884413 + }, + { + "epoch": 2.8523489932885906, + "flow_matching_loss": 0.022363264858722687, + "grad_norm": 0.21401777863502502, + "learning_rate": 2.362304948395906e-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": 0.008863104181364179, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.004238870576955378, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.0219650786370039, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.022363264858722687, + "weighted_rkd_loss": 0.01098253931850195 + }, + { + "epoch": 2.8691275167785237, + "flow_matching_loss": 0.021167028322815896, + "grad_norm": 0.2148161679506302, + "learning_rate": 2.3541684892728157e-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": 0.008788073668256402, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.004058226011693478, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.021634373534470796, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.021167028322815896, + "weighted_rkd_loss": 0.010817186767235398 + }, + { + "epoch": 2.8859060402684564, + "flow_matching_loss": 0.02436531288549304, + "grad_norm": 0.21661841869354248, + "learning_rate": 2.345994666052883e-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": 0.010374712059274316, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.004931548284366727, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.025680972170084714, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.02436531288549304, + "weighted_rkd_loss": 0.012840486085042357 + }, + { + "epoch": 2.902684563758389, + "flow_matching_loss": 0.022530085407197475, + "grad_norm": 0.17632728815078735, + "learning_rate": 2.337783836285614e-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": 0.008637029910460114, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0039325655903667215, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.021206625550985337, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.022530085407197475, + "weighted_rkd_loss": 0.010603312775492668 + }, + { + "epoch": 2.9194630872483223, + "flow_matching_loss": 0.020638199150562288, + "grad_norm": 0.21533295512199402, + "learning_rate": 2.3295363591393026e-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": 0.009914667624980212, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.004476642981171608, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.024305978044867515, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.020638199150562288, + "weighted_rkd_loss": 0.012152989022433757 + }, + { + "epoch": 2.936241610738255, + "flow_matching_loss": 0.022945385053753854, + "grad_norm": 0.24226795136928558, + "learning_rate": 2.3212525953853174e-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": 0.009643034590408207, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.005037153465673327, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.024323222693055868, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.022945385053753854, + "weighted_rkd_loss": 0.012161611346527934 + }, + { + "epoch": 2.953020134228188, + "flow_matching_loss": 0.021384773962199687, + "grad_norm": 0.20606893301010132, + "learning_rate": 2.3129329073823206e-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": 0.008418774185702204, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0036228498327545823, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.020460398215800524, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.021384773962199687, + "weighted_rkd_loss": 0.010230199107900262 + }, + { + "epoch": 2.969798657718121, + "flow_matching_loss": 0.02141566090285778, + "grad_norm": 0.21554461121559143, + "learning_rate": 2.3045776590604175e-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": 0.010217978153377771, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.004908383940346539, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.025344339944422246, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.02141566090285778, + "weighted_rkd_loss": 0.012672169972211123 + }, + { + "epoch": 2.9865771812080535, + "flow_matching_loss": 0.02188427047803998, + "grad_norm": 0.23835872113704681, + "learning_rate": 2.296187215905237e-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": 0.008762435987591743, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.00426900724414736, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.021793879009783267, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.02188427047803998, + "weighted_rkd_loss": 0.010896939504891633 + }, + { + "epoch": 3.0033557046979866, + "flow_matching_loss": 0.022038573306053875, + "grad_norm": 0.1795431524515152, + "learning_rate": 2.2877619449419437e-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": 0.007520042033866048, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.003554393211379647, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.018594477139413357, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.022038573306053875, + "weighted_rkd_loss": 0.009297238569706679 + }, + { + "epoch": 3.0201342281879193, + "flow_matching_loss": 0.02141010984778404, + "grad_norm": 0.2697470486164093, + "learning_rate": 2.2793022147191843e-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": 0.009075135085731744, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.004234813805669546, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.022385083977133034, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.02141010984778404, + "weighted_rkd_loss": 0.011192541988566517 + }, + { + "epoch": 3.0369127516778525, + "flow_matching_loss": 0.02096653329208493, + "grad_norm": 0.18201664090156555, + "learning_rate": 2.2708083952929634e-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": 0.009220260940492154, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.004256098275072872, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.022696620132774115, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.02096653329208493, + "weighted_rkd_loss": 0.011348310066387058 + }, + { + "epoch": 3.053691275167785, + "flow_matching_loss": 0.022351188492029904, + "grad_norm": 0.2090246081352234, + "learning_rate": 2.2622808582104594e-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": 0.00882717128843069, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0034880529856309295, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.02114239577203989, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.022351188492029904, + "weighted_rkd_loss": 0.010571197886019946 + }, + { + "epoch": 3.0704697986577183, + "flow_matching_loss": 0.020570892561227083, + "grad_norm": 0.18636099994182587, + "learning_rate": 2.2537199764937703e-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": 0.009009950887411832, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.004132049717009067, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.02215195130556822, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.020570892561227083, + "weighted_rkd_loss": 0.01107597565278411 + }, + { + "epoch": 3.087248322147651, + "flow_matching_loss": 0.02188956905156374, + "grad_norm": 0.18344570696353912, + "learning_rate": 2.245126124623595e-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": 0.008544707810506225, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.00419144555926323, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.021280860900878905, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.02188956905156374, + "weighted_rkd_loss": 0.010640430450439452 + }, + { + "epoch": 3.1040268456375837, + "flow_matching_loss": 0.022426276095211507, + "grad_norm": 0.2003953754901886, + "learning_rate": 2.2364996785228547e-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": 0.008748158859089018, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.003919238899834454, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.02141555678099394, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.022426276095211507, + "weighted_rkd_loss": 0.01070777839049697 + }, + { + "epoch": 3.120805369127517, + "flow_matching_loss": 0.02033795341849327, + "grad_norm": 0.22842928767204285, + "learning_rate": 2.227841015540248e-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": 0.007575100101530552, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0034250223077833654, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01857522251084447, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.02033795341849327, + "weighted_rkd_loss": 0.009287611255422235 + }, + { + "epoch": 3.1375838926174495, + "flow_matching_loss": 0.020844181906431912, + "grad_norm": 0.2263597548007965, + "learning_rate": 2.2191505144337433e-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": 0.007685949653387069, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.003535176650620997, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.018907075840979816, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.020844181906431912, + "weighted_rkd_loss": 0.009453537920489908 + }, + { + "epoch": 3.1543624161073827, + "flow_matching_loss": 0.019808660540729762, + "grad_norm": 0.18585313856601715, + "learning_rate": 2.210428555354013e-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": 0.009581155283376575, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.004591489001177252, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.023753799498081207, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.019808660540729762, + "weighted_rkd_loss": 0.011876899749040604 + }, + { + "epoch": 3.1711409395973154, + "flow_matching_loss": 0.020991289615631105, + "grad_norm": 0.22622573375701904, + "learning_rate": 2.201675519827802e-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": 0.009574074717238545, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.004285450396127999, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.023433599807322025, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.020991289615631105, + "weighted_rkd_loss": 0.011716799903661012 + }, + { + "epoch": 3.1879194630872485, + "flow_matching_loss": 0.02278821486979723, + "grad_norm": 0.22967985272407532, + "learning_rate": 2.1928917907412403e-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": 0.009182234713807701, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0036421044263988734, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.022006573714315893, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.02278821486979723, + "weighted_rkd_loss": 0.011003286857157946 + }, + { + "epoch": 3.204697986577181, + "flow_matching_loss": 0.021890855114907026, + "grad_norm": 0.1822008639574051, + "learning_rate": 2.1840777523230937e-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": 0.008652022993192076, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.004090635036118328, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.02139468090608716, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.021890855114907026, + "weighted_rkd_loss": 0.01069734045304358 + }, + { + "epoch": 3.221476510067114, + "flow_matching_loss": 0.02066585961729288, + "grad_norm": 0.1782473474740982, + "learning_rate": 2.1752337901279552e-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": 0.00783657394349575, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.003442372544668615, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.019115520268678667, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.02066585961729288, + "weighted_rkd_loss": 0.009557760134339333 + }, + { + "epoch": 3.238255033557047, + "flow_matching_loss": 0.020921053178608418, + "grad_norm": 0.21969790756702423, + "learning_rate": 2.166360291019383e-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": 0.009407800249755382, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0038952767383307217, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.02271087756380439, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.020921053178608418, + "weighted_rkd_loss": 0.011355438781902195 + }, + { + "epoch": 3.2550335570469797, + "flow_matching_loss": 0.01944717802107334, + "grad_norm": 0.19404731690883636, + "learning_rate": 2.157457643152973e-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": 0.00837903949432075, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0037874118657782673, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.020545491110533477, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.01944717802107334, + "weighted_rkd_loss": 0.010272745555266739 + }, + { + "epoch": 3.271812080536913, + "flow_matching_loss": 0.0224097715690732, + "grad_norm": 0.17556366324424744, + "learning_rate": 2.148526235959385e-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": 0.008008029777556658, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0038434152025729417, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01985947471112013, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.0224097715690732, + "weighted_rkd_loss": 0.009929737355560065 + }, + { + "epoch": 3.2885906040268456, + "flow_matching_loss": 0.019902830384671687, + "grad_norm": 0.2150319367647171, + "learning_rate": 2.1395664601273016e-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": 0.008881382131949066, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0041137875756248835, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.021876551769673826, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.019902830384671687, + "weighted_rkd_loss": 0.010938275884836913 + }, + { + "epoch": 3.3053691275167787, + "flow_matching_loss": 0.020638061966747046, + "grad_norm": 0.20082297921180725, + "learning_rate": 2.130578707586345e-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": 0.009990708529949188, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.004830724210478365, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.024812141060829164, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.020638061966747046, + "weighted_rkd_loss": 0.012406070530414582 + }, + { + "epoch": 3.3221476510067114, + "flow_matching_loss": 0.021972791384905578, + "grad_norm": 0.36457595229148865, + "learning_rate": 2.1215633714899264e-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": 0.008448011986911298, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.004079858679324389, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.020975882653146982, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.021972791384905578, + "weighted_rkd_loss": 0.010487941326573491 + }, + { + "epoch": 3.3389261744966445, + "flow_matching_loss": 0.020128148049116133, + "grad_norm": 0.2117772102355957, + "learning_rate": 2.112520846198053e-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": 0.009189584525302052, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.003930398868396879, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.022309568151831626, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.020128148049116133, + "weighted_rkd_loss": 0.011154784075915813 + }, + { + "epoch": 3.3557046979865772, + "flow_matching_loss": 0.02076246850192547, + "grad_norm": 0.22179049253463745, + "learning_rate": 2.103451527260074e-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": 0.0077252968214452265, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0034601765451952814, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.018910769931972028, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.02076246850192547, + "weighted_rkd_loss": 0.009455384965986014 + }, + { + "epoch": 3.37248322147651, + "flow_matching_loss": 0.019000662211328744, + "grad_norm": 0.22441506385803223, + "learning_rate": 2.094355811397381e-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": 0.007931182719767093, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0037250424269586802, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01958740781992674, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.019000662211328744, + "weighted_rkd_loss": 0.00979370390996337 + }, + { + "epoch": 3.389261744966443, + "flow_matching_loss": 0.02324447510764003, + "grad_norm": 0.18134300410747528, + "learning_rate": 2.0852340964860522e-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": 0.007788776373490691, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0033203586237505077, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.018897911440581082, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.02324447510764003, + "weighted_rkd_loss": 0.009448955720290541 + }, + { + "epoch": 3.4060402684563758, + "flow_matching_loss": 0.02150107007473707, + "grad_norm": 0.20962178707122803, + "learning_rate": 2.076086781539447e-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": 0.0074673259630799295, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0032554172910749914, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.018190069310367107, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.02150107007473707, + "weighted_rkd_loss": 0.009095034655183553 + }, + { + "epoch": 3.422818791946309, + "flow_matching_loss": 0.018979523424059154, + "grad_norm": 0.2038407325744629, + "learning_rate": 2.066914266690755e-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": 0.008069745684042572, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0033081255503930153, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.019447616953402757, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.018979523424059154, + "weighted_rkd_loss": 0.009723808476701378 + }, + { + "epoch": 3.4395973154362416, + "flow_matching_loss": 0.018519976176321507, + "grad_norm": 0.2189130187034607, + "learning_rate": 2.0577169531754906e-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": 0.008641075901687145, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.003642701031640172, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.02092485288158059, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.018519976176321507, + "weighted_rkd_loss": 0.010462426440790295 + }, + { + "epoch": 3.4563758389261743, + "flow_matching_loss": 0.023320261761546134, + "grad_norm": 0.19270910322666168, + "learning_rate": 2.048495243313942e-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": 0.006468802364543081, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.003018935560248792, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.015956540312618017, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.023320261761546134, + "weighted_rkd_loss": 0.007978270156309008 + }, + { + "epoch": 3.4731543624161074, + "flow_matching_loss": 0.01975588295608759, + "grad_norm": 0.1822284758090973, + "learning_rate": 2.039249540493572e-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": 0.008690228406339883, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.00411738664843142, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.02149784378707409, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.01975588295608759, + "weighted_rkd_loss": 0.010748921893537045 + }, + { + "epoch": 3.48993288590604, + "flow_matching_loss": 0.01874214904382825, + "grad_norm": 0.1515883505344391, + "learning_rate": 2.0299802491513755e-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": 0.008525742217898369, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0038174554938450456, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.020868940092623234, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.01874214904382825, + "weighted_rkd_loss": 0.010434470046311617 + }, + { + "epoch": 3.5067114093959733, + "flow_matching_loss": 0.018373808544129133, + "grad_norm": 0.22102390229701996, + "learning_rate": 2.0206877747561828e-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": 0.007662647403776646, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0035474081989377735, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.018872702866792677, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.018373808544129133, + "weighted_rkd_loss": 0.009436351433396339 + }, + { + "epoch": 3.523489932885906, + "flow_matching_loss": 0.017856397293508054, + "grad_norm": 0.15369464457035065, + "learning_rate": 2.011372523790927e-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": 0.008157176710665227, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0034173124935477973, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.019731666054576637, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.017856397293508054, + "weighted_rkd_loss": 0.009865833027288318 + }, + { + "epoch": 3.540268456375839, + "flow_matching_loss": 0.01998805645853281, + "grad_norm": 0.1758951097726822, + "learning_rate": 2.0020349037348643e-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": 0.006989141134545207, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0031998127698898314, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.017178095038980247, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.01998805645853281, + "weighted_rkd_loss": 0.008589047519490123 + }, + { + "epoch": 3.557046979865772, + "flow_matching_loss": 0.018932559806853533, + "grad_norm": 0.17690835893154144, + "learning_rate": 1.9926753230457443e-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": 0.007896357215940952, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0036027796799317003, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.0193954941816628, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.018932559806853533, + "weighted_rkd_loss": 0.0096977470908314 + }, + { + "epoch": 3.573825503355705, + "flow_matching_loss": 0.023601194098591804, + "grad_norm": 0.19478943943977356, + "learning_rate": 1.983294191141948e-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": 0.009023548383265734, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.004056708514690399, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.02210380509495735, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.023601194098591804, + "weighted_rkd_loss": 0.011051902547478676 + }, + { + "epoch": 3.5906040268456376, + "flow_matching_loss": 0.020795693341642617, + "grad_norm": 0.15999910235404968, + "learning_rate": 1.973891918384575e-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": 0.008007427817210555, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.003208912722766399, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01922376835718751, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.020795693341642617, + "weighted_rkd_loss": 0.009611884178593755 + }, + { + "epoch": 3.6073825503355703, + "flow_matching_loss": 0.021222106367349624, + "grad_norm": 0.21778394281864166, + "learning_rate": 1.9644689160594945e-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": 0.00793103752657771, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.003410175652243197, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01927225077524781, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.021222106367349624, + "weighted_rkd_loss": 0.009636125387623906 + }, + { + "epoch": 3.6241610738255035, + "flow_matching_loss": 0.017535629961639643, + "grad_norm": 0.21843908727169037, + "learning_rate": 1.9550255963593553e-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": 0.007544249249622226, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0029765219427645207, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.018065020348876713, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.017535629961639643, + "weighted_rkd_loss": 0.009032510174438357 + }, + { + "epoch": 3.640939597315436, + "flow_matching_loss": 0.019194155745208264, + "grad_norm": 0.20880937576293945, + "learning_rate": 1.9455623723655524e-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": 0.00752829727716744, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0031787700252607467, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.018235364742577077, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.019194155745208264, + "weighted_rkd_loss": 0.009117682371288538 + }, + { + "epoch": 3.6577181208053693, + "flow_matching_loss": 0.019468952622264623, + "grad_norm": 0.18948209285736084, + "learning_rate": 1.9360796580301606e-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": 0.007364970911294222, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0032614518189802767, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.017991393711417912, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.019468952622264623, + "weighted_rkd_loss": 0.008995696855708956 + }, + { + "epoch": 3.674496644295302, + "flow_matching_loss": 0.022077844850718974, + "grad_norm": 0.16150549054145813, + "learning_rate": 1.9265778681578245e-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": 0.00765900406986475, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0032386504113674163, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01855665845796466, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.022077844850718974, + "weighted_rkd_loss": 0.00927832922898233 + }, + { + "epoch": 3.6912751677852347, + "flow_matching_loss": 0.019230139348655938, + "grad_norm": 0.25615647435188293, + "learning_rate": 1.9170574183876143e-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": 0.007380595942959189, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0032563595566898586, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01801755167543888, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.019230139348655938, + "weighted_rkd_loss": 0.00900877583771944 + }, + { + "epoch": 3.708053691275168, + "flow_matching_loss": 0.017652079928666353, + "grad_norm": 0.1817786544561386, + "learning_rate": 1.9075187251748452e-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": 0.007373360265046358, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0032471208134666084, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.017993841134011746, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.017652079928666353, + "weighted_rkd_loss": 0.008996920567005873 + }, + { + "epoch": 3.7248322147651005, + "flow_matching_loss": 0.01961641861125827, + "grad_norm": 0.17127038538455963, + "learning_rate": 1.8979622057728596e-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": 0.00780273643322289, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.003444134397432208, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.019049607403576373, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.01961641861125827, + "weighted_rkd_loss": 0.009524803701788187 + }, + { + "epoch": 3.7416107382550337, + "flow_matching_loss": 0.019033155217766762, + "grad_norm": 0.17839007079601288, + "learning_rate": 1.8883882782147748e-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": 0.00856158221140504, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.003942649811506271, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.02106581414118409, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.019033155217766762, + "weighted_rkd_loss": 0.010532907070592045 + }, + { + "epoch": 3.7583892617449663, + "flow_matching_loss": 0.019040965475142002, + "grad_norm": 0.1941649317741394, + "learning_rate": 1.8787973612951967e-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": 0.007480817148461938, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0031261443509720266, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01808777851983905, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.019040965475142002, + "weighted_rkd_loss": 0.009043889259919525 + }, + { + "epoch": 3.7751677852348995, + "flow_matching_loss": 0.019748223666101693, + "grad_norm": 0.1636851280927658, + "learning_rate": 1.8691898745519028e-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": 0.008198972046375274, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.003771033138036728, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.02016897704452276, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.019748223666101693, + "weighted_rkd_loss": 0.01008448852226138 + }, + { + "epoch": 3.791946308724832, + "flow_matching_loss": 0.022198669519275425, + "grad_norm": 0.22628392279148102, + "learning_rate": 1.8595662382474863e-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": 0.007077188324183225, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0031255503417924047, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.017279926873743533, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.022198669519275425, + "weighted_rkd_loss": 0.008639963436871767 + }, + { + "epoch": 3.8087248322147653, + "flow_matching_loss": 0.016530673112720252, + "grad_norm": 0.21448732912540436, + "learning_rate": 1.8499268733509766e-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": 0.007735572382807731, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0035640710266306996, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.019035216141492128, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.016530673112720252, + "weighted_rkd_loss": 0.009517608070746064 + }, + { + "epoch": 3.825503355704698, + "flow_matching_loss": 0.021276569552719594, + "grad_norm": 0.14651033282279968, + "learning_rate": 1.84027220151942e-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": 0.00765763008967042, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0031805066973902285, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01849576672539115, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.021276569552719594, + "weighted_rkd_loss": 0.009247883362695575 + }, + { + "epoch": 3.8422818791946307, + "flow_matching_loss": 0.021208910923451185, + "grad_norm": 0.1972508579492569, + "learning_rate": 1.83060264507944e-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": 0.007044157804921269, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0029331859084777536, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.017021501343697308, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.021208910923451185, + "weighted_rkd_loss": 0.008510750671848654 + }, + { + "epoch": 3.859060402684564, + "flow_matching_loss": 0.018153464980423452, + "grad_norm": 0.1617676019668579, + "learning_rate": 1.82091862700876e-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": 0.008048683824017643, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.003274902165867388, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.019372269883751868, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.018153464980423452, + "weighted_rkd_loss": 0.009686134941875934 + }, + { + "epoch": 3.8758389261744965, + "flow_matching_loss": 0.019316170644015074, + "grad_norm": 0.20088458061218262, + "learning_rate": 1.8112205709177023e-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": 0.00764038790948689, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0031374069745652377, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01841818243265152, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.019316170644015074, + "weighted_rkd_loss": 0.00920909121632576 + }, + { + "epoch": 3.8926174496644297, + "flow_matching_loss": 0.01878544017672539, + "grad_norm": 0.17331726849079132, + "learning_rate": 1.8015089010306585e-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": 0.006989720230922103, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0030413487227633597, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.0170207891613245, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.01878544017672539, + "weighted_rkd_loss": 0.00851039458066225 + }, + { + "epoch": 3.9093959731543624, + "flow_matching_loss": 0.018327188584953545, + "grad_norm": 0.18315960466861725, + "learning_rate": 1.7917840421675298e-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": 0.006432220642454922, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0026728694792836905, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.015537310857325792, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.018327188584953545, + "weighted_rkd_loss": 0.007768655428662896 + }, + { + "epoch": 3.926174496644295, + "flow_matching_loss": 0.017242155969142914, + "grad_norm": 0.17186187207698822, + "learning_rate": 1.7820464197251474e-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": 0.006151645421050489, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0027259318390861153, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.015029222704470158, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.017242155969142914, + "weighted_rkd_loss": 0.007514611352235079 + }, + { + "epoch": 3.942953020134228, + "flow_matching_loss": 0.02237462541088462, + "grad_norm": 0.24216608703136444, + "learning_rate": 1.7722964596586627e-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": 0.007923036813735962, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.003261329559609294, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01910740341991186, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.02237462541088462, + "weighted_rkd_loss": 0.00955370170995593 + }, + { + "epoch": 3.959731543624161, + "flow_matching_loss": 0.018751779105514288, + "grad_norm": 0.23497825860977173, + "learning_rate": 1.7625345884629144e-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": 0.007246668357402087, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0031278782756999136, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01762121506035328, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.018751779105514288, + "weighted_rkd_loss": 0.00881060753017664 + }, + { + "epoch": 3.976510067114094, + "flow_matching_loss": 0.01858077123761177, + "grad_norm": 0.20044608414173126, + "learning_rate": 1.7527612331537722e-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": 0.006888248072937131, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.002850493707228452, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.016626989655196666, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.01858077123761177, + "weighted_rkd_loss": 0.008313494827598333 + }, + { + "epoch": 3.9932885906040267, + "flow_matching_loss": 0.019617028534412384, + "grad_norm": 0.19277773797512054, + "learning_rate": 1.7429768212494593e-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": 0.007408017152920365, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.00341418138705194, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.0182302157394588, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.019617028534412384, + "weighted_rkd_loss": 0.0091151078697294 + }, + { + "epoch": 4.010067114093959, + "flow_matching_loss": 0.01653716042637825, + "grad_norm": 0.16038158535957336, + "learning_rate": 1.7331817807518497e-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": 0.006630103243514896, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.003022903867531568, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01628311024978757, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.01653716042637825, + "weighted_rkd_loss": 0.008141555124893785 + }, + { + "epoch": 4.026845637583893, + "flow_matching_loss": 0.018175854813307524, + "grad_norm": 0.19850464165210724, + "learning_rate": 1.723376540127745e-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": 0.00760634196922183, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0034310010029003023, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.0186436852440238, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.018175854813307524, + "weighted_rkd_loss": 0.0093218426220119 + }, + { + "epoch": 4.043624161073826, + "flow_matching_loss": 0.019848077557981014, + "grad_norm": 0.16854286193847656, + "learning_rate": 1.7135615282901357e-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": 0.008194463048130273, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.00404913816601038, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.020438064355403185, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.019848077557981014, + "weighted_rkd_loss": 0.010219032177701592 + }, + { + "epoch": 4.060402684563758, + "flow_matching_loss": 0.020925424993038177, + "grad_norm": 0.20957912504673004, + "learning_rate": 1.7037371745794358e-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": 0.006662683980539441, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0029504644102416933, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.016275832336395978, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.020925424993038177, + "weighted_rkd_loss": 0.008137916168197989 + }, + { + "epoch": 4.077181208053691, + "flow_matching_loss": 0.019630491454154252, + "grad_norm": 0.1673985868692398, + "learning_rate": 1.6939039087447026e-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": 0.006573117896914482, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.002810140186920762, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01595637621358037, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.019630491454154252, + "weighted_rkd_loss": 0.007978188106790185 + }, + { + "epoch": 4.093959731543624, + "flow_matching_loss": 0.021586596686393022, + "grad_norm": 0.201747328042984, + "learning_rate": 1.6840621609248397e-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": 0.007233408000320196, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.003126069717109203, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01759288590401411, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.021586596686393022, + "weighted_rkd_loss": 0.008796442952007055 + }, + { + "epoch": 4.110738255033557, + "flow_matching_loss": 0.018460760451853277, + "grad_norm": 0.2036897987127304, + "learning_rate": 1.6742123616297794e-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": 0.007287621404975653, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.003143525216728449, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.017718768026679755, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.018460760451853277, + "weighted_rkd_loss": 0.008859384013339877 + }, + { + "epoch": 4.12751677852349, + "flow_matching_loss": 0.017445159237831832, + "grad_norm": 0.16198286414146423, + "learning_rate": 1.664354941721651e-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": 0.006410072930157185, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.003017538005951792, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.0158376838080585, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.017445159237831832, + "weighted_rkd_loss": 0.00791884190402925 + }, + { + "epoch": 4.144295302013423, + "flow_matching_loss": 0.016915331315249204, + "grad_norm": 0.16699963808059692, + "learning_rate": 1.6544903323959355e-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": 0.005673299496993423, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0020931367063894866, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.013439735677093268, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.016915331315249204, + "weighted_rkd_loss": 0.006719867838546634 + }, + { + "epoch": 4.1610738255033555, + "flow_matching_loss": 0.020208620838820933, + "grad_norm": 0.215232715010643, + "learning_rate": 1.6446189651626023e-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": 0.006692374125123024, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0026895080460235476, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01607425631955266, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.020208620838820933, + "weighted_rkd_loss": 0.00803712815977633 + }, + { + "epoch": 4.177852348993288, + "flow_matching_loss": 0.020926382672041655, + "grad_norm": 0.22336435317993164, + "learning_rate": 1.6347412718272316e-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": 0.007076423778198659, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.003159638389479369, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01731248591095209, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.020926382672041655, + "weighted_rkd_loss": 0.008656242955476045 + }, + { + "epoch": 4.194630872483222, + "flow_matching_loss": 0.016524071246385573, + "grad_norm": 0.1518680602312088, + "learning_rate": 1.624857684472131e-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": 0.006714969547465444, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.003003756864927709, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01643369598314166, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.016524071246385573, + "weighted_rkd_loss": 0.00821684799157083 + }, + { + "epoch": 4.2114093959731544, + "flow_matching_loss": 0.018874516151845454, + "grad_norm": 0.17943860590457916, + "learning_rate": 1.6149686354374283e-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": 0.006851326674222946, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.002800596947781742, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.0165032503195107, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.018874516151845454, + "weighted_rkd_loss": 0.00825162515975535 + }, + { + "epoch": 4.228187919463087, + "flow_matching_loss": 0.016614831425249575, + "grad_norm": 0.25215211510658264, + "learning_rate": 1.605074557302165e-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": 0.006883031688630581, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0028047516010701657, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01657081479206681, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.016614831425249575, + "weighted_rkd_loss": 0.008285407396033406 + }, + { + "epoch": 4.24496644295302, + "flow_matching_loss": 0.01997323101386428, + "grad_norm": 0.20624391734600067, + "learning_rate": 1.59517588286537e-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": 0.006685473956167698, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.00293945602606982, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01631040396168828, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.01997323101386428, + "weighted_rkd_loss": 0.00815520198084414 + }, + { + "epoch": 4.261744966442953, + "flow_matching_loss": 0.02013480234891176, + "grad_norm": 0.1708594560623169, + "learning_rate": 1.5852730451271324e-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": 0.005825302423909306, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.002814886881969869, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.014465491846203804, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.02013480234891176, + "weighted_rkd_loss": 0.007232745923101902 + }, + { + "epoch": 4.278523489932886, + "flow_matching_loss": 0.017857098765671254, + "grad_norm": 0.16984955966472626, + "learning_rate": 1.5753664772696546e-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": 0.006671967078000307, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0027174580260179936, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01606139214709401, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.017857098765671254, + "weighted_rkd_loss": 0.008030696073547006 + }, + { + "epoch": 4.295302013422819, + "flow_matching_loss": 0.017334267776459456, + "grad_norm": 0.15617665648460388, + "learning_rate": 1.565456612638307e-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": 0.006426696386188269, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.002855884563177824, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.015709277335554363, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.017334267776459456, + "weighted_rkd_loss": 0.007854638667777181 + }, + { + "epoch": 4.3120805369127515, + "flow_matching_loss": 0.019305859599262475, + "grad_norm": 0.16668863594532013, + "learning_rate": 1.5555438847226732e-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": 0.00716956895776093, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0026791229262016713, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.017018260527402164, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.019305859599262475, + "weighted_rkd_loss": 0.008509130263701082 + }, + { + "epoch": 4.328859060402684, + "flow_matching_loss": 0.019607679545879365, + "grad_norm": 0.1785610169172287, + "learning_rate": 1.545628727137584e-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": 0.009049271885305642, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0035267358762212098, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.021625279542058708, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.019607679545879365, + "weighted_rkd_loss": 0.010812639771029354 + }, + { + "epoch": 4.345637583892618, + "flow_matching_loss": 0.018800497893244027, + "grad_norm": 0.1957562118768692, + "learning_rate": 1.535711573604153e-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": 0.00803911597467959, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.003370839380659163, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.019449071399867533, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.018800497893244027, + "weighted_rkd_loss": 0.009724535699933767 + }, + { + "epoch": 4.3624161073825505, + "flow_matching_loss": 0.017709657829254866, + "grad_norm": 0.2177783101797104, + "learning_rate": 1.5257928579308032e-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": 0.006389908026903868, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0026443781214766206, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.015424194186925888, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.017709657829254866, + "weighted_rkd_loss": 0.007712097093462944 + }, + { + "epoch": 4.379194630872483, + "flow_matching_loss": 0.01953520327806473, + "grad_norm": 0.22895203530788422, + "learning_rate": 1.5158730139942902e-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": 0.008311857981607319, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.003941608686000108, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.02056532446295023, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.01953520327806473, + "weighted_rkd_loss": 0.010282662231475115 + }, + { + "epoch": 4.395973154362416, + "flow_matching_loss": 0.020596037432551385, + "grad_norm": 0.13484011590480804, + "learning_rate": 1.5059524757207239e-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": 0.005220752139575779, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0022500797058455646, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01269158385694027, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.020596037432551385, + "weighted_rkd_loss": 0.006345791928470135 + }, + { + "epoch": 4.412751677852349, + "flow_matching_loss": 0.018211192917078733, + "grad_norm": 0.17946451902389526, + "learning_rate": 1.4960316770665866e-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": 0.006453538197092712, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0030796703533269463, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.015986746829003095, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.018211192917078733, + "weighted_rkd_loss": 0.007993373414501547 + }, + { + "epoch": 4.429530201342282, + "flow_matching_loss": 0.0171947973780334, + "grad_norm": 0.1605665385723114, + "learning_rate": 1.4861110519997495e-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": 0.007121807662770152, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0029179507866501806, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.017161566112190486, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.0171947973780334, + "weighted_rkd_loss": 0.008580783056095243 + }, + { + "epoch": 4.446308724832215, + "flow_matching_loss": 0.017047136928886174, + "grad_norm": 0.21463190019130707, + "learning_rate": 1.4761910344804913e-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": 0.006163895828649402, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0027821387513540683, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01510993055999279, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.017047136928886174, + "weighted_rkd_loss": 0.007554965279996395 + }, + { + "epoch": 4.4630872483221475, + "flow_matching_loss": 0.019098941888660192, + "grad_norm": 0.13163383305072784, + "learning_rate": 1.466272058442516e-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": 0.005966320005245507, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0023630001582205296, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.0142956402618438, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.019098941888660192, + "weighted_rkd_loss": 0.0071478201309219 + }, + { + "epoch": 4.47986577181208, + "flow_matching_loss": 0.01778299594298005, + "grad_norm": 0.2292172610759735, + "learning_rate": 1.4563545577739676e-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": 0.006542552821338176, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0030771528254263103, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01616225838661194, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.01778299594298005, + "weighted_rkd_loss": 0.00808112919330597 + }, + { + "epoch": 4.496644295302014, + "flow_matching_loss": 0.015495003666728735, + "grad_norm": 0.17960447072982788, + "learning_rate": 1.4464389662984543e-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": 0.007274207659065723, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.002846747334115207, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.017395162582397462, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.015495003666728735, + "weighted_rkd_loss": 0.008697581291198731 + }, + { + "epoch": 4.5134228187919465, + "flow_matching_loss": 0.016512521728873252, + "grad_norm": 0.18700137734413147, + "learning_rate": 1.4365257177560684e-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": 0.006523366179317236, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0025913154473528265, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.015638047736138104, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.016512521728873252, + "weighted_rkd_loss": 0.007819023868069052 + }, + { + "epoch": 4.530201342281879, + "flow_matching_loss": 0.018928809091448784, + "grad_norm": 0.1533912718296051, + "learning_rate": 1.426615245784416e-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": 0.006431228248402476, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0026095522684045137, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.015472008753567934, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.018928809091448784, + "weighted_rkd_loss": 0.007736004376783967 + }, + { + "epoch": 4.546979865771812, + "flow_matching_loss": 0.01595159135758877, + "grad_norm": 0.13736169040203094, + "learning_rate": 1.4167079838996468e-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": 0.006889982311986387, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0030951215769164262, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01687508588656783, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.01595159135758877, + "weighted_rkd_loss": 0.008437542943283916 + }, + { + "epoch": 4.563758389261745, + "flow_matching_loss": 0.017109850142151117, + "grad_norm": 0.1820959895849228, + "learning_rate": 1.4068043654774897e-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": 0.006259463052265346, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0027393482509069146, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01525827432051301, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.017109850142151117, + "weighted_rkd_loss": 0.007629137160256505 + }, + { + "epoch": 4.580536912751678, + "flow_matching_loss": 0.018156492337584495, + "grad_norm": 0.182673841714859, + "learning_rate": 1.3969048237342992e-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": 0.0066155354026705025, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0028239200590178372, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01605499079450965, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.018156492337584495, + "weighted_rkd_loss": 0.008027495397254825 + }, + { + "epoch": 4.597315436241611, + "flow_matching_loss": 0.01978831607848406, + "grad_norm": 0.20100091397762299, + "learning_rate": 1.3870097917081011e-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": 0.0070770631544291975, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.002857843821402639, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.017011970281600952, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.01978831607848406, + "weighted_rkd_loss": 0.008505985140800476 + }, + { + "epoch": 4.614093959731544, + "flow_matching_loss": 0.0186252910643816, + "grad_norm": 0.18295343220233917, + "learning_rate": 1.3771197022396508e-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": 0.006896005268208683, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.002676388737745583, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.016468399157747626, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.0186252910643816, + "weighted_rkd_loss": 0.008234199578873813 + }, + { + "epoch": 4.630872483221476, + "flow_matching_loss": 0.017001602426171304, + "grad_norm": 0.18725721538066864, + "learning_rate": 1.3672349879535015e-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": 0.0075272177811712025, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.00305442416574806, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01810885975137353, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.017001602426171304, + "weighted_rkd_loss": 0.009054429875686765 + }, + { + "epoch": 4.64765100671141, + "flow_matching_loss": 0.01720100613310933, + "grad_norm": 0.15701332688331604, + "learning_rate": 1.3573560812390781e-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": 0.006909766187891364, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0026912813424132763, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.016510813497006893, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.01720100613310933, + "weighted_rkd_loss": 0.008255406748503447 + }, + { + "epoch": 4.6644295302013425, + "flow_matching_loss": 0.017884473502635955, + "grad_norm": 0.14792506396770477, + "learning_rate": 1.3474834142317635e-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": 0.006335597159340978, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.002531678543891758, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.015202872920781374, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.017884473502635955, + "weighted_rkd_loss": 0.007601436460390687 + }, + { + "epoch": 4.681208053691275, + "flow_matching_loss": 0.01976926876232028, + "grad_norm": 0.17041365802288055, + "learning_rate": 1.3376174187939955e-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": 0.0057981258258223535, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.002652096108067781, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01424834793433547, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.01976926876232028, + "weighted_rkd_loss": 0.007124173967167735 + }, + { + "epoch": 4.697986577181208, + "flow_matching_loss": 0.019026348367333412, + "grad_norm": 0.15966573357582092, + "learning_rate": 1.3277585264963761e-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": 0.0060622354969382284, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0023348112939856947, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.014459282439202071, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.019026348367333412, + "weighted_rkd_loss": 0.0072296412196010355 + }, + { + "epoch": 4.714765100671141, + "flow_matching_loss": 0.017377084400504828, + "grad_norm": 0.22258582711219788, + "learning_rate": 1.3179071685987928e-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": 0.0056399278808385135, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.002916206861846149, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.014196062553673982, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.017377084400504828, + "weighted_rkd_loss": 0.007098031276836991 + }, + { + "epoch": 4.731543624161074, + "flow_matching_loss": 0.017920501250773668, + "grad_norm": 0.2010204941034317, + "learning_rate": 1.308063776031554e-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": 0.005940242158249021, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.002402682788670063, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.014283167012035847, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.017920501250773668, + "weighted_rkd_loss": 0.007141583506017924 + }, + { + "epoch": 4.748322147651007, + "flow_matching_loss": 0.01623355569317937, + "grad_norm": 0.17003734409809113, + "learning_rate": 1.2982287793765389e-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": 0.005624843528494239, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0024171904427930714, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.013666877429932356, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.01623355569317937, + "weighted_rkd_loss": 0.006833438714966178 + }, + { + "epoch": 4.76510067114094, + "flow_matching_loss": 0.016382606141269206, + "grad_norm": 0.16437426209449768, + "learning_rate": 1.2884026088483616e-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": 0.005247332667931914, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0022283923462964594, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.012723057717084884, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.016382606141269206, + "weighted_rkd_loss": 0.006361528858542442 + }, + { + "epoch": 4.781879194630872, + "flow_matching_loss": 0.020086504332721234, + "grad_norm": 0.20742954313755035, + "learning_rate": 1.2785856942755543e-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": 0.006500152428634465, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0027116223354823887, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.015711927181109787, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.020086504332721234, + "weighted_rkd_loss": 0.007855963590554894 + }, + { + "epoch": 4.798657718120805, + "flow_matching_loss": 0.020468197017908096, + "grad_norm": 0.18006539344787598, + "learning_rate": 1.2687784650817616e-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": 0.006521564652211964, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0030275082564912735, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.016070637572556733, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.020468197017908096, + "weighted_rkd_loss": 0.008035318786278367 + }, + { + "epoch": 4.815436241610739, + "flow_matching_loss": 0.020298132486641408, + "grad_norm": 0.21230530738830566, + "learning_rate": 1.2589813502669603e-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": 0.006211647461168468, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.002518127963412553, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.014941422827541828, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.020298132486641408, + "weighted_rkd_loss": 0.007470711413770914 + }, + { + "epoch": 4.832214765100671, + "flow_matching_loss": 0.018210323061794043, + "grad_norm": 0.17627596855163574, + "learning_rate": 1.2491947783886907e-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": 0.004903101362287998, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0021681559388525783, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.011974358558654785, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.018210323061794043, + "weighted_rkd_loss": 0.005987179279327392 + }, + { + "epoch": 4.848993288590604, + "flow_matching_loss": 0.016133444383740425, + "grad_norm": 0.13107307255268097, + "learning_rate": 1.2394191775433106e-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": 0.005635178997181356, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.002382622950244695, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.013652980979532003, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.016133444383740425, + "weighted_rkd_loss": 0.006826490489766001 + }, + { + "epoch": 4.865771812080537, + "flow_matching_loss": 0.016500497329980136, + "grad_norm": 0.1938703954219818, + "learning_rate": 1.2296549753472696e-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": 0.005910794297233224, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.002559423819184303, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.014381012320518494, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.016500497329980136, + "weighted_rkd_loss": 0.007190506160259247 + }, + { + "epoch": 4.882550335570469, + "flow_matching_loss": 0.015015983954071999, + "grad_norm": 0.18545067310333252, + "learning_rate": 1.2199025989184032e-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": 0.005890693236142397, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0025859495741315185, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.014367335941642523, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.015015983954071999, + "weighted_rkd_loss": 0.007183667970821261 + }, + { + "epoch": 4.899328859060403, + "flow_matching_loss": 0.015698899049311876, + "grad_norm": 0.20254436135292053, + "learning_rate": 1.21016247485725e-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": 0.0064388867001980545, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0024498644983395935, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01532763782888651, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.015698899049311876, + "weighted_rkd_loss": 0.007663818914443255 + }, + { + "epoch": 4.916107382550336, + "flow_matching_loss": 0.016660439129918814, + "grad_norm": 0.18414856493473053, + "learning_rate": 1.2004350292283897e-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": 0.006194327911362052, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0027924057794734836, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.015181061439216137, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.016660439129918814, + "weighted_rkd_loss": 0.007590530719608068 + }, + { + "epoch": 4.932885906040268, + "flow_matching_loss": 0.01687433701008558, + "grad_norm": 0.16476738452911377, + "learning_rate": 1.1907206875418068e-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": 0.006046470603905618, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0024416908156126736, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.014534632116556168, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.01687433701008558, + "weighted_rkd_loss": 0.007267316058278084 + }, + { + "epoch": 4.949664429530201, + "flow_matching_loss": 0.016659815981984138, + "grad_norm": 0.15063215792179108, + "learning_rate": 1.1810198747342769e-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": 0.005068508721888065, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0021296532708220185, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.012266670726239682, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.016659815981984138, + "weighted_rkd_loss": 0.006133335363119841 + }, + { + "epoch": 4.966442953020135, + "flow_matching_loss": 0.014483388233929872, + "grad_norm": 0.18998196721076965, + "learning_rate": 1.1713330151507786e-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": 0.006280959723517299, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0027005726704373957, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.015262492187321185, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.014483388233929872, + "weighted_rkd_loss": 0.007631246093660593 + }, + { + "epoch": 4.983221476510067, + "flow_matching_loss": 0.016974644828587772, + "grad_norm": 0.1717025190591812, + "learning_rate": 1.161660532525931e-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": 0.006291400594636798, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0028575255651958286, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.015440326556563378, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.016974644828587772, + "weighted_rkd_loss": 0.007720163278281689 + }, + { + "epoch": 5.0, + "flow_matching_loss": 0.015258622635155917, + "grad_norm": 0.16946367919445038, + "learning_rate": 1.1520028499654584e-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": 0.00629035490565002, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.002330243296455592, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.014910953119397163, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.015258622635155917, + "weighted_rkd_loss": 0.007455476559698582 + }, + { + "epoch": 5.016778523489933, + "flow_matching_loss": 0.017898726742714643, + "grad_norm": 0.14455287158489227, + "learning_rate": 1.1423603899276833e-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": 0.0053604997927322985, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.002361709764227271, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01308270925655961, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.017898726742714643, + "weighted_rkd_loss": 0.006541354628279805 + }, + { + "epoch": 5.033557046979865, + "flow_matching_loss": 0.01671588756144047, + "grad_norm": 0.18688088655471802, + "learning_rate": 1.1327335742050442e-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": 0.005779878236353397, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.002441789919976145, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01400154661387205, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.01671588756144047, + "weighted_rkd_loss": 0.007000773306936025 + }, + { + "epoch": 5.050335570469799, + "flow_matching_loss": 0.015817246027290822, + "grad_norm": 0.1989966332912445, + "learning_rate": 1.1231228239056467e-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": 0.006527934223413467, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.002765071636531502, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.015820940118283035, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.015817246027290822, + "weighted_rkd_loss": 0.007910470059141517 + }, + { + "epoch": 5.067114093959732, + "flow_matching_loss": 0.01937075946480036, + "grad_norm": 0.19405128061771393, + "learning_rate": 1.113528559434845e-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": 0.006799411680549383, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0029540552641265094, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01655287845060229, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.01937075946480036, + "weighted_rkd_loss": 0.008276439225301146 + }, + { + "epoch": 5.083892617449664, + "flow_matching_loss": 0.01867743730545044, + "grad_norm": 0.13398496806621552, + "learning_rate": 1.1039512004768476e-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": 0.006300268415361643, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.003146882087457925, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.015747418720275165, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.01867743730545044, + "weighted_rkd_loss": 0.007873709360137583 + }, + { + "epoch": 5.100671140939597, + "flow_matching_loss": 0.016951379459351303, + "grad_norm": 0.20637381076812744, + "learning_rate": 1.0943911659763612e-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": 0.0054731245152652265, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0021580980042926966, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.013104346767067909, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.016951379459351303, + "weighted_rkd_loss": 0.006552173383533954 + }, + { + "epoch": 5.117449664429531, + "flow_matching_loss": 0.01617958638817072, + "grad_norm": 0.17529357969760895, + "learning_rate": 1.084848874120266e-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": 0.004983928962610662, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.002211737749166787, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.012179595697671176, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.01617958638817072, + "weighted_rkd_loss": 0.006089797848835588 + }, + { + "epoch": 5.134228187919463, + "flow_matching_loss": 0.01798154441639781, + "grad_norm": 0.1679459512233734, + "learning_rate": 1.0753247423193196e-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": 0.0051181812305003405, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.002054891409352422, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.012291253916919232, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.01798154441639781, + "weighted_rkd_loss": 0.006145626958459616 + }, + { + "epoch": 5.151006711409396, + "flow_matching_loss": 0.016323540639132262, + "grad_norm": 0.20896075665950775, + "learning_rate": 1.0658191871899012e-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": 0.005931211682036519, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0027073875186033546, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.014569810871034861, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.016323540639132262, + "weighted_rkd_loss": 0.007284905435517431 + }, + { + "epoch": 5.167785234899329, + "flow_matching_loss": 0.018675233144313096, + "grad_norm": 0.21605846285820007, + "learning_rate": 1.0563326245357852e-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": 0.005888878158293665, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0024224749999120832, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.014200231153517961, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.018675233144313096, + "weighted_rkd_loss": 0.007100115576758981 + }, + { + "epoch": 5.184563758389261, + "flow_matching_loss": 0.015594534948468209, + "grad_norm": 0.21472957730293274, + "learning_rate": 1.0468654693299557e-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": 0.005332688824273646, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.002128972625359893, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.012794350273907185, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.015594534948468209, + "weighted_rkd_loss": 0.006397175136953592 + }, + { + "epoch": 5.201342281879195, + "flow_matching_loss": 0.01680131210014224, + "grad_norm": 0.15244846045970917, + "learning_rate": 1.0374181356964501e-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": 0.005860675545409322, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.002612880081869662, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.014334231335669755, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.01680131210014224, + "weighted_rkd_loss": 0.007167115667834878 + }, + { + "epoch": 5.218120805369128, + "flow_matching_loss": 0.01609828118234873, + "grad_norm": 0.17836423218250275, + "learning_rate": 1.0279910368922465e-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": 0.005648004543036223, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.002437116380315274, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.013733125384896993, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.01609828118234873, + "weighted_rkd_loss": 0.006866562692448497 + }, + { + "epoch": 5.23489932885906, + "flow_matching_loss": 0.016079106274992226, + "grad_norm": 0.1312175989151001, + "learning_rate": 1.018584585289186e-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": 0.006313463742844761, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.002551958546973765, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.015178885776549578, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.016079106274992226, + "weighted_rkd_loss": 0.007589442888274789 + }, + { + "epoch": 5.251677852348993, + "flow_matching_loss": 0.018048496171832084, + "grad_norm": 0.14827987551689148, + "learning_rate": 1.0091991923559348e-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": 0.006192484707571566, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.002579770388547331, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.014964739698916674, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.018048496171832084, + "weighted_rkd_loss": 0.007482369849458337 + }, + { + "epoch": 5.268456375838926, + "flow_matching_loss": 0.016098976973444222, + "grad_norm": 0.21780844032764435, + "learning_rate": 9.99835268639984e-06, + "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": 0.005791219044476747, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0024931206949986517, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.014075558725744486, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.016098976973444222, + "weighted_rkd_loss": 0.007037779362872243 + }, + { + "epoch": 5.285234899328859, + "flow_matching_loss": 0.0159193379804492, + "grad_norm": 0.19609591364860535, + "learning_rate": 9.904932237496913e-06, + "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": 0.004937235242687165, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.00216703173937276, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.012041502259671688, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.0159193379804492, + "weighted_rkd_loss": 0.006020751129835844 + }, + { + "epoch": 5.302013422818792, + "flow_matching_loss": 0.014781306311488152, + "grad_norm": 0.2759782373905182, + "learning_rate": 9.811734663363659e-06, + "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": 0.005513566290028393, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.00228801624616608, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.013315148837864398, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.014781306311488152, + "weighted_rkd_loss": 0.006657574418932199 + }, + { + "epoch": 5.318791946308725, + "flow_matching_loss": 0.01723882174119353, + "grad_norm": 0.16407190263271332, + "learning_rate": 9.718764040763875e-06, + "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": 0.0059700444340705875, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0022358012502081692, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.014175890199840068, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.01723882174119353, + "weighted_rkd_loss": 0.007087945099920034 + }, + { + "epoch": 5.3355704697986575, + "flow_matching_loss": 0.016165380273014306, + "grad_norm": 0.15735647082328796, + "learning_rate": 9.626024436533777e-06, + "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": 0.005720328283496201, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.002354632946662605, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.013795289304107427, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.016165380273014306, + "weighted_rkd_loss": 0.006897644652053714 + }, + { + "epoch": 5.35234899328859, + "flow_matching_loss": 0.01835033930838108, + "grad_norm": 0.18436121940612793, + "learning_rate": 9.533519907404095e-06, + "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": 0.005160203762352466, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.002100703300675377, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.012421110738068819, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.01835033930838108, + "weighted_rkd_loss": 0.006210555369034409 + }, + { + "epoch": 5.369127516778524, + "flow_matching_loss": 0.0171617632266134, + "grad_norm": 0.17887401580810547, + "learning_rate": 9.4412544998226e-06, + "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": 0.004801495070569217, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0022160978172905744, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.011819087900221348, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.0171617632266134, + "weighted_rkd_loss": 0.005909543950110674 + }, + { + "epoch": 5.385906040268456, + "flow_matching_loss": 0.01628105537965894, + "grad_norm": 0.21056130528450012, + "learning_rate": 9.349232249777124e-06, + "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": 0.0059770388295874, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0024258324061520397, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01437990991398692, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.01628105537965894, + "weighted_rkd_loss": 0.00718995495699346 + }, + { + "epoch": 5.402684563758389, + "flow_matching_loss": 0.019007640797644852, + "grad_norm": 0.2918168008327484, + "learning_rate": 9.257457182618985e-06, + "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": 0.005413648206740618, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.002455686731263995, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.013282982911914588, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.019007640797644852, + "weighted_rkd_loss": 0.006641491455957294 + }, + { + "epoch": 5.419463087248322, + "flow_matching_loss": 0.015169256553053855, + "grad_norm": 0.15156346559524536, + "learning_rate": 9.165933312886944e-06, + "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": 0.004989207885228097, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.002060519380029291, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01203893506899476, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.015169256553053855, + "weighted_rkd_loss": 0.00601946753449738 + }, + { + "epoch": 5.436241610738255, + "flow_matching_loss": 0.016222430299967526, + "grad_norm": 0.153106227517128, + "learning_rate": 9.07466464413155e-06, + "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": 0.0059154522605240345, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0021968859364278615, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.014027790259569883, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.016222430299967526, + "weighted_rkd_loss": 0.0070138951297849415 + }, + { + "epoch": 5.453020134228188, + "flow_matching_loss": 0.016991760395467282, + "grad_norm": 0.23132218420505524, + "learning_rate": 8.983655168740048e-06, + "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": 0.006680940859951079, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.002693852235097438, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.016055733896791935, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.016991760395467282, + "weighted_rkd_loss": 0.008027866948395967 + }, + { + "epoch": 5.469798657718121, + "flow_matching_loss": 0.016313020419329406, + "grad_norm": 0.14859110116958618, + "learning_rate": 8.892908867761718e-06, + "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": 0.006743181962519884, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.002991057641338557, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.016477421671152116, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.016313020419329406, + "weighted_rkd_loss": 0.008238710835576058 + }, + { + "epoch": 5.4865771812080535, + "flow_matching_loss": 0.018052593991160392, + "grad_norm": 0.16553567349910736, + "learning_rate": 8.802429710733745e-06, + "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": 0.004876571684144437, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.002291589602828026, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.012044732971116901, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.018052593991160392, + "weighted_rkd_loss": 0.006022366485558451 + }, + { + "epoch": 5.503355704697986, + "flow_matching_loss": 0.01667319508269429, + "grad_norm": 0.18746444582939148, + "learning_rate": 8.71222165550757e-06, + "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": 0.004696449846960604, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0018371382844634353, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01123003801330924, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.01667319508269429, + "weighted_rkd_loss": 0.00561501900665462 + }, + { + "epoch": 5.52013422818792, + "flow_matching_loss": 0.015698192361742257, + "grad_norm": 0.19901099801063538, + "learning_rate": 8.622288648075748e-06, + "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": 0.0054225688334554436, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0021200297283940016, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.012965167313814164, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.015698192361742257, + "weighted_rkd_loss": 0.006482583656907082 + }, + { + "epoch": 5.5369127516778525, + "flow_matching_loss": 0.016813407838344573, + "grad_norm": 0.16385048627853394, + "learning_rate": 8.532634622399391e-06, + "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": 0.004744720063172281, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.002088817674666643, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.011578257894143462, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.016813407838344573, + "weighted_rkd_loss": 0.005789128947071731 + }, + { + "epoch": 5.553691275167785, + "flow_matching_loss": 0.015239028632640839, + "grad_norm": 0.2184711992740631, + "learning_rate": 8.44326350023601e-06, + "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": 0.005157306534238159, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0025373138021677734, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.012851926730945706, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.015239028632640839, + "weighted_rkd_loss": 0.006425963365472853 + }, + { + "epoch": 5.570469798657718, + "flow_matching_loss": 0.015581510215997695, + "grad_norm": 0.15516401827335358, + "learning_rate": 8.354179190968002e-06, + "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": 0.004385465546511114, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0015736499801278115, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01034458102658391, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.015581510215997695, + "weighted_rkd_loss": 0.005172290513291955 + }, + { + "epoch": 5.587248322147651, + "flow_matching_loss": 0.018718068767338992, + "grad_norm": 0.21937350928783417, + "learning_rate": 8.265385591431661e-06, + "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": 0.005961455311626196, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.002444499370176345, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.014367409981787205, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.018718068767338992, + "weighted_rkd_loss": 0.007183704990893602 + }, + { + "epoch": 5.604026845637584, + "flow_matching_loss": 0.014946687500923872, + "grad_norm": 0.14376555383205414, + "learning_rate": 8.176886585746671e-06, + "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": 0.004589359811507166, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0019512920989654958, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.011130011733621359, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.014946687500923872, + "weighted_rkd_loss": 0.005565005866810679 + }, + { + "epoch": 5.620805369127517, + "flow_matching_loss": 0.013927073497325183, + "grad_norm": 0.1809721142053604, + "learning_rate": 8.088686045146247e-06, + "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": 0.005342445382848382, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0022515961551107465, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.012936487048864364, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.013927073497325183, + "weighted_rkd_loss": 0.006468243524432182 + }, + { + "epoch": 5.6375838926174495, + "flow_matching_loss": 0.015200373250991106, + "grad_norm": 0.19390113651752472, + "learning_rate": 8.00078782780775e-06, + "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": 0.005480418261140585, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.002286845410708338, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01324768178164959, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.015200373250991106, + "weighted_rkd_loss": 0.006623840890824795 + }, + { + "epoch": 5.654362416107382, + "flow_matching_loss": 0.017947501968592405, + "grad_norm": 0.1948995292186737, + "learning_rate": 7.913195778683979e-06, + "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": 0.00509554436430335, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.002246081584598869, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.012437170092016458, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.017947501968592405, + "weighted_rkd_loss": 0.006218585046008229 + }, + { + "epoch": 5.671140939597316, + "flow_matching_loss": 0.01727897087112069, + "grad_norm": 0.19190214574337006, + "learning_rate": 7.825913729334908e-06, + "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": 0.005547965504229068, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.00244416092755273, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.013540091924369335, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.01727897087112069, + "weighted_rkd_loss": 0.006770045962184668 + }, + { + "epoch": 5.6879194630872485, + "flow_matching_loss": 0.018374356906861067, + "grad_norm": 0.1317068636417389, + "learning_rate": 7.73894549776013e-06, + "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": 0.0055965505540370945, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0022137959022074938, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.013406896963715553, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.018374356906861067, + "weighted_rkd_loss": 0.006703448481857777 + }, + { + "epoch": 5.704697986577181, + "flow_matching_loss": 0.015805857069790364, + "grad_norm": 0.2368166297674179, + "learning_rate": 7.65229488823184e-06, + "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": 0.0053893422009423375, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0023492945125326513, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.013127978704869748, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.015805857069790364, + "weighted_rkd_loss": 0.006563989352434874 + }, + { + "epoch": 5.721476510067114, + "flow_matching_loss": 0.01789247877895832, + "grad_norm": 0.16481158137321472, + "learning_rate": 7.565965691128396e-06, + "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": 0.004735127929598093, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0020541234756819905, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.011524379393085838, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.01789247877895832, + "weighted_rkd_loss": 0.005762189696542919 + }, + { + "epoch": 5.7382550335570475, + "flow_matching_loss": 0.01610357007011771, + "grad_norm": 0.19617056846618652, + "learning_rate": 7.47996168276856e-06, + "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": 0.004507807712070644, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0017629609443247317, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01077857632189989, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.01610357007011771, + "weighted_rkd_loss": 0.005389288160949945 + }, + { + "epoch": 5.75503355704698, + "flow_matching_loss": 0.016361732501536608, + "grad_norm": 0.16310381889343262, + "learning_rate": 7.394286625246258e-06, + "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": 0.005316798621788621, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0018169037997722627, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.012450501043349504, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.016361732501536608, + "weighted_rkd_loss": 0.006225250521674752 + }, + { + "epoch": 5.771812080536913, + "flow_matching_loss": 0.01591941826045513, + "grad_norm": 0.14360399544239044, + "learning_rate": 7.308944266266063e-06, + "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": 0.0050241583026945594, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.00202472114469856, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.012073037773370742, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.01591941826045513, + "weighted_rkd_loss": 0.006036518886685371 + }, + { + "epoch": 5.7885906040268456, + "flow_matching_loss": 0.0171136362478137, + "grad_norm": 0.25523585081100464, + "learning_rate": 7.223938338979233e-06, + "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": 0.0050462506711483005, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.001889492047484964, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.011981993541121483, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.0171136362478137, + "weighted_rkd_loss": 0.005990996770560741 + }, + { + "epoch": 5.805369127516778, + "flow_matching_loss": 0.017496591899544, + "grad_norm": 0.1567636877298355, + "learning_rate": 7.139272561820397e-06, + "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": 0.00436295464169234, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0016373030026443302, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.010363212414085865, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.017496591899544, + "weighted_rkd_loss": 0.005181606207042932 + }, + { + "epoch": 5.822147651006711, + "flow_matching_loss": 0.014734391588717699, + "grad_norm": 0.15370847284793854, + "learning_rate": 7.054950638344936e-06, + "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": 0.0047149185324087736, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0020497960387729107, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01147963311523199, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.014734391588717699, + "weighted_rkd_loss": 0.005739816557615995 + }, + { + "epoch": 5.8389261744966445, + "flow_matching_loss": 0.016384201217442752, + "grad_norm": 0.18166224658489227, + "learning_rate": 6.970976257066941e-06, + "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": 0.005584712466225028, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.002454860834404826, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.013624285720288754, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.016384201217442752, + "weighted_rkd_loss": 0.006812142860144377 + }, + { + "epoch": 5.855704697986577, + "flow_matching_loss": 0.015255410224199295, + "grad_norm": 0.22488172352313995, + "learning_rate": 6.887353091297896e-06, + "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": 0.004990561841987073, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0019109268672764302, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.011892050690948962, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.015255410224199295, + "weighted_rkd_loss": 0.005946025345474481 + }, + { + "epoch": 5.87248322147651, + "flow_matching_loss": 0.016748811397701503, + "grad_norm": 0.17458610236644745, + "learning_rate": 6.804084798985965e-06, + "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": 0.005321139958687127, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.001955470233224332, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01259775012731552, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.016748811397701503, + "weighted_rkd_loss": 0.00629887506365776 + }, + { + "epoch": 5.889261744966443, + "flow_matching_loss": 0.017467737291008235, + "grad_norm": 0.2232614904642105, + "learning_rate": 6.721175022556009e-06, + "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": 0.005391786922700703, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.002220256079453975, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.013003829959779977, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.017467737291008235, + "weighted_rkd_loss": 0.006501914979889989 + }, + { + "epoch": 5.906040268456376, + "flow_matching_loss": 0.015664891712367533, + "grad_norm": 0.1693728268146515, + "learning_rate": 6.638627388750245e-06, + "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": 0.005543821305036545, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.001970905240159482, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01305854767560959, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.015664891712367533, + "weighted_rkd_loss": 0.006529273837804795 + }, + { + "epoch": 5.922818791946309, + "flow_matching_loss": 0.018745915219187736, + "grad_norm": 0.1922639161348343, + "learning_rate": 6.556445508469583e-06, + "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": 0.004745194385759532, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0019202875206246971, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.011410676408559084, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.018745915219187736, + "weighted_rkd_loss": 0.005705338204279542 + }, + { + "epoch": 5.939597315436242, + "flow_matching_loss": 0.017241082154214382, + "grad_norm": 0.19612009823322296, + "learning_rate": 6.474632976615708e-06, + "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": 0.005357312108390033, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.002517049363814294, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.013231673557311297, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.017241082154214382, + "weighted_rkd_loss": 0.006615836778655649 + }, + { + "epoch": 5.956375838926174, + "flow_matching_loss": 0.016150518599897624, + "grad_norm": 0.17087864875793457, + "learning_rate": 6.393193371933791e-06, + "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": 0.005487588234245777, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0025250311009585857, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.013500207662582397, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.016150518599897624, + "weighted_rkd_loss": 0.0067501038312911986 + }, + { + "epoch": 5.973154362416107, + "flow_matching_loss": 0.015010569430887698, + "grad_norm": 0.14347855746746063, + "learning_rate": 6.312130256855977e-06, + "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": 0.004399256804026664, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0019206363591365515, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.010719149839133024, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.015010569430887698, + "weighted_rkd_loss": 0.005359574919566512 + }, + { + "epoch": 5.989932885906041, + "flow_matching_loss": 0.015253768581897021, + "grad_norm": 0.1432712972164154, + "learning_rate": 6.23144717734552e-06, + "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": 0.005114776152186096, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0019915621145628394, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.012221114430576562, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.015253768581897021, + "weighted_rkd_loss": 0.006110557215288281 + }, + { + "epoch": 6.006711409395973, + "flow_matching_loss": 0.015888039860874413, + "grad_norm": 0.21380585432052612, + "learning_rate": 6.151147662741695e-06, + "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": 0.004904805589467287, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0020907162339426576, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.011900327261537313, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.015888039860874413, + "weighted_rkd_loss": 0.0059501636307686566 + }, + { + "epoch": 6.023489932885906, + "flow_matching_loss": 0.01624783258885145, + "grad_norm": 0.17643627524375916, + "learning_rate": 6.071235225605415e-06, + "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": 0.004626808851025998, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0019845670787617562, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01123818475753069, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.01624783258885145, + "weighted_rkd_loss": 0.005619092378765345 + }, + { + "epoch": 6.040268456375839, + "flow_matching_loss": 0.016807893011718988, + "grad_norm": 0.13113495707511902, + "learning_rate": 5.991713361565546e-06, + "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": 0.00433010421693325, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0018857970600947738, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.010546005656942725, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.016807893011718988, + "weighted_rkd_loss": 0.005273002828471362 + }, + { + "epoch": 6.057046979865772, + "flow_matching_loss": 0.014842659141868353, + "grad_norm": 0.2173433154821396, + "learning_rate": 5.912585549166051e-06, + "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": 0.005738632497377693, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0024297324125654995, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.013906997302547098, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.014842659141868353, + "weighted_rkd_loss": 0.006953498651273549 + }, + { + "epoch": 6.073825503355705, + "flow_matching_loss": 0.015946427546441554, + "grad_norm": 0.17731747031211853, + "learning_rate": 5.833855249713777e-06, + "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": 0.0042516337241977455, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0017034933145623655, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.010206760745495558, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.015946427546441554, + "weighted_rkd_loss": 0.005103380372747779 + }, + { + "epoch": 6.090604026845638, + "flow_matching_loss": 0.017258873023092747, + "grad_norm": 0.18392245471477509, + "learning_rate": 5.7555259071270645e-06, + "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": 0.005527050793170929, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0023105093743652104, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.013364610914140939, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.017258873023092747, + "weighted_rkd_loss": 0.0066823054570704695 + }, + { + "epoch": 6.10738255033557, + "flow_matching_loss": 0.014458374958485365, + "grad_norm": 0.15555499494075775, + "learning_rate": 5.677600947785117e-06, + "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": 0.004228003416210413, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0018116092891432346, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.010267616109922528, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.014458374958485365, + "weighted_rkd_loss": 0.005133808054961264 + }, + { + "epoch": 6.124161073825503, + "flow_matching_loss": 0.014426036737859249, + "grad_norm": 0.12407287210226059, + "learning_rate": 5.600083780378099e-06, + "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": 0.004010814358480275, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0015088674030266703, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.009530496131628752, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.014426036737859249, + "weighted_rkd_loss": 0.004765248065814376 + }, + { + "epoch": 6.140939597315437, + "flow_matching_loss": 0.015246859192848206, + "grad_norm": 0.19507645070552826, + "learning_rate": 5.522977795758038e-06, + "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": 0.005159773584455252, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0018581768497824668, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.012177724111825227, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.015246859192848206, + "weighted_rkd_loss": 0.0060888620559126135 + }, + { + "epoch": 6.157718120805369, + "flow_matching_loss": 0.016874602250754833, + "grad_norm": 0.1572439819574356, + "learning_rate": 5.4462863667904805e-06, + "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": 0.005385365919210017, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0023309690644964577, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.013101700972765685, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.016874602250754833, + "weighted_rkd_loss": 0.006550850486382842 + }, + { + "epoch": 6.174496644295302, + "flow_matching_loss": 0.017885672207921744, + "grad_norm": 0.17745526134967804, + "learning_rate": 5.370012848206983e-06, + "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": 0.0054987040115520355, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.002123111695982516, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01312051983550191, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.017885672207921744, + "weighted_rkd_loss": 0.006560259917750955 + }, + { + "epoch": 6.191275167785235, + "flow_matching_loss": 0.014684625109657646, + "grad_norm": 0.2061356008052826, + "learning_rate": 5.29416057645834e-06, + "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": 0.004387064115144312, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0018910664482973516, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.010665194783359765, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.014684625109657646, + "weighted_rkd_loss": 0.005332597391679883 + }, + { + "epoch": 6.208053691275167, + "flow_matching_loss": 0.016030605789273976, + "grad_norm": 0.14315824210643768, + "learning_rate": 5.218732869568635e-06, + "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": 0.0051259903470054265, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.002160561189521104, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.012412541825324298, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.016030605789273976, + "weighted_rkd_loss": 0.006206270912662149 + }, + { + "epoch": 6.224832214765101, + "flow_matching_loss": 0.014568359684199094, + "grad_norm": 0.14933782815933228, + "learning_rate": 5.14373302699013e-06, + "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": 0.0043758470332249996, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0017608381458558142, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.010512532247230411, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.014568359684199094, + "weighted_rkd_loss": 0.005256266123615206 + }, + { + "epoch": 6.241610738255034, + "flow_matching_loss": 0.015898380894213915, + "grad_norm": 0.15506769716739655, + "learning_rate": 5.069164329458894e-06, + "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": 0.004872710234485567, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0019389147055335343, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.011684335116297007, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.015898380894213915, + "weighted_rkd_loss": 0.005842167558148504 + }, + { + "epoch": 6.258389261744966, + "flow_matching_loss": 0.01669842703267932, + "grad_norm": 0.2566884160041809, + "learning_rate": 4.995030038851342e-06, + "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": 0.004822015180252492, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0017544238246046007, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.011398454243317247, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.01669842703267932, + "weighted_rkd_loss": 0.0056992271216586236 + }, + { + "epoch": 6.275167785234899, + "flow_matching_loss": 0.01423853449523449, + "grad_norm": 0.15029947459697723, + "learning_rate": 4.9213333980415e-06, + "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": 0.005033561633899808, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0019928340218029915, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01205995730124414, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.01423853449523449, + "weighted_rkd_loss": 0.00602997865062207 + }, + { + "epoch": 6.291946308724833, + "flow_matching_loss": 0.016005691885948182, + "grad_norm": 0.1524253487586975, + "learning_rate": 4.848077630759175e-06, + "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": 0.004667837568558752, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.002215351990889758, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.011551027372479438, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.016005691885948182, + "weighted_rkd_loss": 0.005775513686239719 + }, + { + "epoch": 6.308724832214765, + "flow_matching_loss": 0.017472335230559112, + "grad_norm": 0.27056488394737244, + "learning_rate": 4.775265941448951e-06, + "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": 0.005071670259349048, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0018874454370234161, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.012030785949900747, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.017472335230559112, + "weighted_rkd_loss": 0.0060153929749503735 + }, + { + "epoch": 6.325503355704698, + "flow_matching_loss": 0.016118133440613747, + "grad_norm": 0.14759497344493866, + "learning_rate": 4.702901515129989e-06, + "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": 0.004513580771163106, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0019892573938705025, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.011016418971121312, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.016118133440613747, + "weighted_rkd_loss": 0.005508209485560656 + }, + { + "epoch": 6.342281879194631, + "flow_matching_loss": 0.015037001110613346, + "grad_norm": 0.2157750427722931, + "learning_rate": 4.630987517256734e-06, + "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": 0.004265101510100067, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.001737450179643929, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.010267653083428741, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.015037001110613346, + "weighted_rkd_loss": 0.005133826541714371 + }, + { + "epoch": 6.359060402684563, + "flow_matching_loss": 0.015858409553766252, + "grad_norm": 0.1426669806241989, + "learning_rate": 4.559527093580412e-06, + "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": 0.00464945612475276, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0019282410736195742, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.011227153288200498, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.015858409553766252, + "weighted_rkd_loss": 0.005613576644100249 + }, + { + "epoch": 6.375838926174497, + "flow_matching_loss": 0.015834339428693056, + "grad_norm": 0.11701840162277222, + "learning_rate": 4.488523370011456e-06, + "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": 0.0038423003163188694, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0015366179984994232, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.00922121866606176, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.015834339428693056, + "weighted_rkd_loss": 0.00461060933303088 + }, + { + "epoch": 6.39261744966443, + "flow_matching_loss": 0.014915316551923751, + "grad_norm": 0.12409483641386032, + "learning_rate": 4.417979452482766e-06, + "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": 0.003954608063213527, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0016807083855383098, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.009589924523606897, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.014915316551923751, + "weighted_rkd_loss": 0.0047949622618034485 + }, + { + "epoch": 6.409395973154362, + "flow_matching_loss": 0.014157744124531746, + "grad_norm": 0.11032804846763611, + "learning_rate": 4.347898426813811e-06, + "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": 0.004619434243068099, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0021324063884094356, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.011371274897828697, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.014157744124531746, + "weighted_rkd_loss": 0.005685637448914349 + }, + { + "epoch": 6.426174496644295, + "flow_matching_loss": 0.014956588950008154, + "grad_norm": 0.14672960340976715, + "learning_rate": 4.278283358575694e-06, + "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": 0.004830200923606753, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0020344901247881355, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.011694891983643175, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.014956588950008154, + "weighted_rkd_loss": 0.005847445991821587 + }, + { + "epoch": 6.442953020134228, + "flow_matching_loss": 0.014917949866503476, + "grad_norm": 0.14469048380851746, + "learning_rate": 4.2091372929570145e-06, + "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": 0.004648088780231774, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.002000366896390915, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.011296544317156076, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.014917949866503476, + "weighted_rkd_loss": 0.005648272158578038 + }, + { + "epoch": 6.459731543624161, + "flow_matching_loss": 0.01866527069360018, + "grad_norm": 0.17728039622306824, + "learning_rate": 4.1404632546306934e-06, + "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": 0.0048375848913565275, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.002021892520133406, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.011697062104940415, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.01866527069360018, + "weighted_rkd_loss": 0.0058485310524702076 + }, + { + "epoch": 6.476510067114094, + "flow_matching_loss": 0.014754966367036104, + "grad_norm": 0.21448470652103424, + "learning_rate": 4.0722642476216286e-06, + "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": 0.00443953056819737, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0017713158624246717, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.010650376975536346, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.014754966367036104, + "weighted_rkd_loss": 0.005325188487768173 + }, + { + "epoch": 6.493288590604027, + "flow_matching_loss": 0.01667445064522326, + "grad_norm": 0.14577926695346832, + "learning_rate": 4.00454325517533e-06, + "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": 0.004969216231256723, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0019304315675981343, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.011868863925337791, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.01667445064522326, + "weighted_rkd_loss": 0.0059344319626688955 + }, + { + "epoch": 6.510067114093959, + "flow_matching_loss": 0.016786788031458853, + "grad_norm": 0.1949937641620636, + "learning_rate": 3.937303239627393e-06, + "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": 0.004295885050669313, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0018998178886249662, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.010491588059812785, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.016786788031458853, + "weighted_rkd_loss": 0.005245794029906392 + }, + { + "epoch": 6.526845637583893, + "flow_matching_loss": 0.01631803922355175, + "grad_norm": 0.21952104568481445, + "learning_rate": 3.870547142273918e-06, + "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": 0.004865309922024608, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.001927953865379095, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.011658573616296053, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.01631803922355175, + "weighted_rkd_loss": 0.005829286808148026 + }, + { + "epoch": 6.543624161073826, + "flow_matching_loss": 0.016452026460319756, + "grad_norm": 0.196889266371727, + "learning_rate": 3.8042778832428736e-06, + "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": 0.004349646950140595, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0019458610797300935, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.010645155049860477, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.016452026460319756, + "weighted_rkd_loss": 0.005322577524930238 + }, + { + "epoch": 6.560402684563758, + "flow_matching_loss": 0.019432486873120068, + "grad_norm": 0.13456007838249207, + "learning_rate": 3.7384983613663195e-06, + "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": 0.004305106052197516, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0016713534423615783, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.010281565645709635, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.019432486873120068, + "weighted_rkd_loss": 0.005140782822854817 + }, + { + "epoch": 6.577181208053691, + "flow_matching_loss": 0.01630993653088808, + "grad_norm": 0.14141227304935455, + "learning_rate": 3.6732114540536503e-06, + "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": 0.0041214486351236704, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.001664979092311114, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.009907876374199986, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.01630993653088808, + "weighted_rkd_loss": 0.004953938187099993 + }, + { + "epoch": 6.593959731543624, + "flow_matching_loss": 0.016547040734440088, + "grad_norm": 0.178508922457695, + "learning_rate": 3.608420017165683e-06, + "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": 0.0038512055296450855, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0015659712022170424, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.009268382238224149, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.016547040734440088, + "weighted_rkd_loss": 0.0046341911191120745 + }, + { + "epoch": 6.610738255033557, + "flow_matching_loss": 0.01753809815272689, + "grad_norm": 0.17572665214538574, + "learning_rate": 3.5441268848897664e-06, + "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": 0.004356299003120512, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0018225304782390594, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.010535128554329276, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.01753809815272689, + "weighted_rkd_loss": 0.005267564277164638 + }, + { + "epoch": 6.62751677852349, + "flow_matching_loss": 0.015104027651250362, + "grad_norm": 0.17830194532871246, + "learning_rate": 3.480334869615792e-06, + "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": 0.004359930544160307, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0016910021542571485, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.010410863300785422, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.015104027651250362, + "weighted_rkd_loss": 0.005205431650392711 + }, + { + "epoch": 6.644295302013423, + "flow_matching_loss": 0.015497284568846225, + "grad_norm": 0.15404456853866577, + "learning_rate": 3.417046761813155e-06, + "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": 0.004788657440803945, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0018028156249783934, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.011380130471661688, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.015497284568846225, + "weighted_rkd_loss": 0.005690065235830844 + }, + { + "epoch": 6.6610738255033555, + "flow_matching_loss": 0.015348309185355902, + "grad_norm": 0.1278257668018341, + "learning_rate": 3.3542653299087263e-06, + "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": 0.004161874623969197, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.001591894228477031, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.009915643278509379, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.015348309185355902, + "weighted_rkd_loss": 0.004957821639254689 + }, + { + "epoch": 6.677852348993289, + "flow_matching_loss": 0.012799481302499771, + "grad_norm": 0.11342970281839371, + "learning_rate": 3.2919933201657117e-06, + "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": 0.004698016052134335, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0017846812726929784, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.011180713446810842, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.012799481302499771, + "weighted_rkd_loss": 0.005590356723405421 + }, + { + "epoch": 6.694630872483222, + "flow_matching_loss": 0.016074532829225063, + "grad_norm": 0.1375473588705063, + "learning_rate": 3.2302334565635584e-06, + "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": 0.004121939861215651, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.001517608918948099, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.009761488577350975, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.016074532829225063, + "weighted_rkd_loss": 0.004880744288675487 + }, + { + "epoch": 6.7114093959731544, + "flow_matching_loss": 0.013632395770400763, + "grad_norm": 0.13031578063964844, + "learning_rate": 3.1689884406787674e-06, + "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": 0.004645679029636085, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0015164404409006237, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.010807798663154244, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.013632395770400763, + "weighted_rkd_loss": 0.005403899331577122 + }, + { + "epoch": 6.728187919463087, + "flow_matching_loss": 0.015095925237983466, + "grad_norm": 0.16176630556583405, + "learning_rate": 3.1082609515667424e-06, + "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": 0.0054400034714490175, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0021193060209043326, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.012999312952160835, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.015095925237983466, + "weighted_rkd_loss": 0.006499656476080418 + }, + { + "epoch": 6.74496644295302, + "flow_matching_loss": 0.01631878176704049, + "grad_norm": 0.1447662115097046, + "learning_rate": 3.048053645644594e-06, + "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": 0.005066341208294034, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0020264827413484453, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.012159165181219577, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.01631878176704049, + "weighted_rkd_loss": 0.0060795825906097885 + }, + { + "epoch": 6.7617449664429525, + "flow_matching_loss": 0.01471923552453518, + "grad_norm": 0.17339105904102325, + "learning_rate": 2.988369156574916e-06, + "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": 0.004106980562210083, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0014705379900988192, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.00968449916690588, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.01471923552453518, + "weighted_rkd_loss": 0.00484224958345294 + }, + { + "epoch": 6.778523489932886, + "flow_matching_loss": 0.01501200720667839, + "grad_norm": 0.13662545382976532, + "learning_rate": 2.9292100951506225e-06, + "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": 0.0036912807961925866, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0015375688206404447, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.008920130459591747, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.01501200720667839, + "weighted_rkd_loss": 0.004460065229795873 + }, + { + "epoch": 6.795302013422819, + "flow_matching_loss": 0.016395939514040947, + "grad_norm": 0.16309326887130737, + "learning_rate": 2.8705790491807033e-06, + "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": 0.00422157181892544, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0016458123573102057, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.010088955983519554, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.016395939514040947, + "weighted_rkd_loss": 0.005044477991759777 + }, + { + "epoch": 6.8120805369127515, + "flow_matching_loss": 0.014857204258441925, + "grad_norm": 0.19027967751026154, + "learning_rate": 2.8124785833770433e-06, + "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": 0.00493489746004343, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0019772169762291014, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.011847011931240558, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.014857204258441925, + "weighted_rkd_loss": 0.005923505965620279 + }, + { + "epoch": 6.828859060402684, + "flow_matching_loss": 0.01854044497013092, + "grad_norm": 0.12597495317459106, + "learning_rate": 2.754911239242241e-06, + "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": 0.004213976627215743, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0018678706022910775, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.010295823868364095, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.01854044497013092, + "weighted_rkd_loss": 0.0051479119341820475 + }, + { + "epoch": 6.845637583892618, + "flow_matching_loss": 0.014754861127585173, + "grad_norm": 0.11598055809736252, + "learning_rate": 2.6978795349584204e-06, + "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": 0.0033723325235769154, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0013575545453932136, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.008102219691500068, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.014754861127585173, + "weighted_rkd_loss": 0.004051109845750034 + }, + { + "epoch": 6.8624161073825505, + "flow_matching_loss": 0.016834757197648285, + "grad_norm": 0.16511546075344086, + "learning_rate": 2.641385965277088e-06, + "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": 0.0039168634917587045, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0017385203973390162, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.009572247229516506, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.016834757197648285, + "weighted_rkd_loss": 0.004786123614758253 + }, + { + "epoch": 6.879194630872483, + "flow_matching_loss": 0.015211101062595845, + "grad_norm": 0.13418087363243103, + "learning_rate": 2.5854330014099923e-06, + "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": 0.004090139199979604, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.001788419799413532, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.00996869821101427, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.015211101062595845, + "weighted_rkd_loss": 0.004984349105507135 + }, + { + "epoch": 6.895973154362416, + "flow_matching_loss": 0.015671815164387225, + "grad_norm": 0.1589498072862625, + "learning_rate": 2.5300230909210298e-06, + "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": 0.004447044571861625, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0016958691005129368, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.010589958261698484, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.015671815164387225, + "weighted_rkd_loss": 0.005294979130849242 + }, + { + "epoch": 6.912751677852349, + "flow_matching_loss": 0.014125759527087212, + "grad_norm": 0.1285037398338318, + "learning_rate": 2.4751586576191983e-06, + "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": 0.003744475147686899, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0014271572756115348, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.008916107751429082, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.014125759527087212, + "weighted_rkd_loss": 0.004458053875714541 + }, + { + "epoch": 6.929530201342282, + "flow_matching_loss": 0.012916694115847348, + "grad_norm": 0.14929188787937164, + "learning_rate": 2.420842101452538e-06, + "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": 0.004445642535574734, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.001718758384231478, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.010610043350607157, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.012916694115847348, + "weighted_rkd_loss": 0.005305021675303579 + }, + { + "epoch": 6.946308724832215, + "flow_matching_loss": 0.013612718507647514, + "grad_norm": 0.14932781457901, + "learning_rate": 2.3670757984031816e-06, + "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": 0.003692518384195864, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0014336197986267508, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.008818656718358398, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.013612718507647514, + "weighted_rkd_loss": 0.004409328359179199 + }, + { + "epoch": 6.9630872483221475, + "flow_matching_loss": 0.013868955336511135, + "grad_norm": 0.1929025948047638, + "learning_rate": 2.313862100383404e-06, + "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": 0.004685251065529883, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0018140813626814634, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01118458341807127, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.013868955336511135, + "weighted_rkd_loss": 0.005592291709035635 + }, + { + "epoch": 6.97986577181208, + "flow_matching_loss": 0.012972024455666541, + "grad_norm": 0.11274044960737228, + "learning_rate": 2.2612033351327406e-06, + "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": 0.003931707842275501, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0015600805985741318, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.009423496341332794, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.012972024455666541, + "weighted_rkd_loss": 0.004711748170666397 + }, + { + "epoch": 6.996644295302014, + "flow_matching_loss": 0.014395527262240648, + "grad_norm": 0.15545383095741272, + "learning_rate": 2.2091018061161788e-06, + "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": 0.004139221063815058, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.001665854500606656, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.009944296535104513, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.014395527262240648, + "weighted_rkd_loss": 0.004972148267552256 + }, + { + "epoch": 7.0134228187919465, + "flow_matching_loss": 0.01504547274671495, + "grad_norm": 0.19873428344726562, + "learning_rate": 2.15755979242338e-06, + "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": 0.00428015049546957, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.001995652518235147, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.010555953532457352, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.01504547274671495, + "weighted_rkd_loss": 0.005277976766228676 + }, + { + "epoch": 7.030201342281879, + "flow_matching_loss": 0.016630757320672273, + "grad_norm": 0.12612350285053253, + "learning_rate": 2.106579548669e-06, + "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": 0.004597616638056934, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0016638765577226878, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.010859109880402684, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.016630757320672273, + "weighted_rkd_loss": 0.005429554940201342 + }, + { + "epoch": 7.046979865771812, + "flow_matching_loss": 0.013459595013409853, + "grad_norm": 0.15308141708374023, + "learning_rate": 2.0561633048940477e-06, + "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": 0.0040692856186069545, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0018304155499208718, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.009968986781314016, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.013459595013409853, + "weighted_rkd_loss": 0.004984493390657008 + }, + { + "epoch": 7.063758389261745, + "flow_matching_loss": 0.016229463275521994, + "grad_norm": 0.15950238704681396, + "learning_rate": 2.006313266468355e-06, + "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": 0.004591330466791987, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0017486129305325448, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.010931274108588695, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.016229463275521994, + "weighted_rkd_loss": 0.005465637054294347 + }, + { + "epoch": 7.080536912751678, + "flow_matching_loss": 0.0157932304777205, + "grad_norm": 0.1525679975748062, + "learning_rate": 1.957031613994097e-06, + "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": 0.0038325243163853883, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0015466227720025926, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.00921167149208486, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.0157932304777205, + "weighted_rkd_loss": 0.00460583574604243 + }, + { + "epoch": 7.097315436241611, + "flow_matching_loss": 0.016353191807866096, + "grad_norm": 0.1053260862827301, + "learning_rate": 1.9083205032103995e-06, + "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": 0.004086917964741588, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0016238337906543166, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.009797669760882855, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.016353191807866096, + "weighted_rkd_loss": 0.004898834880441428 + }, + { + "epoch": 7.114093959731544, + "flow_matching_loss": 0.01574868755415082, + "grad_norm": 0.1536342352628708, + "learning_rate": 1.8601820648990536e-06, + "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": 0.0035363011527806522, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0014526980870869012, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.008525300398468971, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.01574868755415082, + "weighted_rkd_loss": 0.0042626501992344855 + }, + { + "epoch": 7.130872483221476, + "flow_matching_loss": 0.014412057958543301, + "grad_norm": 0.13013885915279388, + "learning_rate": 1.8126184047912936e-06, + "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": 0.003889480617363006, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0015578919497784226, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.009336853120476007, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.014412057958543301, + "weighted_rkd_loss": 0.004668426560238003 + }, + { + "epoch": 7.14765100671141, + "flow_matching_loss": 0.014226209558546542, + "grad_norm": 0.11760559678077698, + "learning_rate": 1.7656316034757024e-06, + "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": 0.0038801743183285, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0018998766783624888, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.009660225315019488, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.014226209558546542, + "weighted_rkd_loss": 0.004830112657509744 + }, + { + "epoch": 7.1644295302013425, + "flow_matching_loss": 0.016309910733252764, + "grad_norm": 0.17850618064403534, + "learning_rate": 1.7192237163071816e-06, + "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": 0.0036488832789473235, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.001311828033067286, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.008609594684094191, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.016309910733252764, + "weighted_rkd_loss": 0.0043047973420470955 + }, + { + "epoch": 7.181208053691275, + "flow_matching_loss": 0.016244333423674106, + "grad_norm": 0.13397976756095886, + "learning_rate": 1.6733967733170573e-06, + "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": 0.004331002850085497, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0017532274941913784, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.010415233299136161, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.016244333423674106, + "weighted_rkd_loss": 0.005207616649568081 + }, + { + "epoch": 7.197986577181208, + "flow_matching_loss": 0.013507804181426764, + "grad_norm": 0.1460564285516739, + "learning_rate": 1.628152779124279e-06, + "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": 0.003937823371961713, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0016270102758426219, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.009502657083794474, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.013507804181426764, + "weighted_rkd_loss": 0.004751328541897237 + }, + { + "epoch": 7.214765100671141, + "flow_matching_loss": 0.014902317523956298, + "grad_norm": 0.12618981301784515, + "learning_rate": 1.583493712847715e-06, + "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": 0.0037965486524626614, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0015084690239746124, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.009101566206663848, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.014902317523956298, + "weighted_rkd_loss": 0.004550783103331924 + }, + { + "epoch": 7.231543624161074, + "flow_matching_loss": 0.016011614352464676, + "grad_norm": 0.14825978875160217, + "learning_rate": 1.5394215280196084e-06, + "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": 0.0037238317308947444, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0015631090383976698, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.009010772593319416, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.016011614352464676, + "weighted_rkd_loss": 0.004505386296659708 + }, + { + "epoch": 7.248322147651007, + "flow_matching_loss": 0.015480771102011204, + "grad_norm": 0.12888647615909576, + "learning_rate": 1.4959381525000904e-06, + "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": 0.004318940313532948, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0016981088323518635, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01033598962239921, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.015480771102011204, + "weighted_rkd_loss": 0.005167994811199605 + }, + { + "epoch": 7.26510067114094, + "flow_matching_loss": 0.01660889098420739, + "grad_norm": 0.12845715880393982, + "learning_rate": 1.453045488392879e-06, + "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": 0.003837773168925196, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0013941763783805073, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.009069722797721624, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.01660889098420739, + "weighted_rkd_loss": 0.004534861398860812 + }, + { + "epoch": 7.281879194630872, + "flow_matching_loss": 0.01536073824390769, + "grad_norm": 0.1266445517539978, + "learning_rate": 1.4107454119620484e-06, + "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": 0.004038789356127381, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0014731727307662368, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.009550751419737935, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.01536073824390769, + "weighted_rkd_loss": 0.0047753757098689675 + }, + { + "epoch": 7.298657718120805, + "flow_matching_loss": 0.015487385308369994, + "grad_norm": 0.1259327232837677, + "learning_rate": 1.3690397735499778e-06, + "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": 0.004094103630632162, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0016427026013843716, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.009830909874290228, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.015487385308369994, + "weighted_rkd_loss": 0.004915454937145114 + }, + { + "epoch": 7.315436241610739, + "flow_matching_loss": 0.016292981710284947, + "grad_norm": 0.208975687623024, + "learning_rate": 1.327930397496403e-06, + "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": 0.0038555726176127793, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0014661048422567546, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.009177249949425459, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.016292981710284947, + "weighted_rkd_loss": 0.004588624974712729 + }, + { + "epoch": 7.332214765100671, + "flow_matching_loss": 0.013602617662400007, + "grad_norm": 0.11648474633693695, + "learning_rate": 1.287419082058604e-06, + "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": 0.004054751852527261, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0015656212577596306, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.009675125079229473, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.013602617662400007, + "weighted_rkd_loss": 0.004837562539614737 + }, + { + "epoch": 7.348993288590604, + "flow_matching_loss": 0.015651093889027835, + "grad_norm": 0.12388869374990463, + "learning_rate": 1.2475075993327589e-06, + "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": 0.003548008017241955, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.001444374053971842, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.008540389966219664, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.015651093889027835, + "weighted_rkd_loss": 0.004270194983109832 + }, + { + "epoch": 7.365771812080537, + "flow_matching_loss": 0.013896638620644807, + "grad_norm": 0.13266593217849731, + "learning_rate": 1.208197695176413e-06, + "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": 0.004024438234046102, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0016310391831211746, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.00967991566285491, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.013896638620644807, + "weighted_rkd_loss": 0.004839957831427455 + }, + { + "epoch": 7.382550335570469, + "flow_matching_loss": 0.013670449983328581, + "grad_norm": 0.17349274456501007, + "learning_rate": 1.1694910891321259e-06, + "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": 0.004115662141703069, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.001552202249877155, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.009783526510000229, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.013670449983328581, + "weighted_rkd_loss": 0.004891763255000114 + }, + { + "epoch": 7.399328859060403, + "flow_matching_loss": 0.014937188103795051, + "grad_norm": 0.23631052672863007, + "learning_rate": 1.1313894743522363e-06, + "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": 0.0042281194357201455, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.001866805658210069, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.010323044611141086, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.014937188103795051, + "weighted_rkd_loss": 0.005161522305570543 + }, + { + "epoch": 7.416107382550336, + "flow_matching_loss": 0.015254345163702965, + "grad_norm": 0.12517230212688446, + "learning_rate": 1.0938945175248016e-06, + "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": 0.003667901479639113, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.001594919990748167, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.008930722950026393, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.015254345163702965, + "weighted_rkd_loss": 0.0044653614750131965 + }, + { + "epoch": 7.432885906040268, + "flow_matching_loss": 0.01722643505781889, + "grad_norm": 0.1511530727148056, + "learning_rate": 1.0570078588007086e-06, + "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": 0.004720710683614016, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0017709346022456885, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.011212356062605977, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.01722643505781889, + "weighted_rkd_loss": 0.005606178031302989 + }, + { + "epoch": 7.449664429530201, + "flow_matching_loss": 0.014523876365274191, + "grad_norm": 0.2035868763923645, + "learning_rate": 1.0207311117219009e-06, + "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": 0.004571941145695746, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0018188763991929592, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01096275863237679, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.014523876365274191, + "weighted_rkd_loss": 0.005481379316188395 + }, + { + "epoch": 7.466442953020135, + "flow_matching_loss": 0.013871367648243904, + "grad_norm": 0.1734429895877838, + "learning_rate": 9.850658631508198e-07, + "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": 0.003869479591958225, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0015727715042885392, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.009311730740591884, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.013871367648243904, + "weighted_rkd_loss": 0.004655865370295942 + }, + { + "epoch": 7.483221476510067, + "flow_matching_loss": 0.0138609291985631, + "grad_norm": 0.16242648661136627, + "learning_rate": 9.500136732009751e-07, + "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": 0.004720749869011343, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0016571854008361696, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01109868511557579, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.0138609291985631, + "weighted_rkd_loss": 0.005549342557787895 + }, + { + "epoch": 7.5, + "flow_matching_loss": 0.013707988150417804, + "grad_norm": 0.12876121699810028, + "learning_rate": 9.155760751687054e-07, + "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": 0.003897508908994496, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.001529338606633246, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.009324356354773045, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.013707988150417804, + "weighted_rkd_loss": 0.0046621781773865225 + }, + { + "epoch": 7.516778523489933, + "flow_matching_loss": 0.014499800279736519, + "grad_norm": 0.16918422281742096, + "learning_rate": 8.817545754661155e-07, + "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": 0.004376136627979576, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0016412930912338198, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.010393566358834505, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.014499800279736519, + "weighted_rkd_loss": 0.005196783179417253 + }, + { + "epoch": 7.533557046979865, + "flow_matching_loss": 0.014267795532941819, + "grad_norm": 0.16523607075214386, + "learning_rate": 8.485506535551668e-07, + "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": 0.004032965656369925, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0016446026973426342, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.009710533916950226, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.014267795532941819, + "weighted_rkd_loss": 0.004855266958475113 + }, + { + "epoch": 7.550335570469799, + "flow_matching_loss": 0.01365661434829235, + "grad_norm": 0.1421973556280136, + "learning_rate": 8.159657618829685e-07, + "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": 0.0037587011931464074, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0014382767374627292, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.008955679135397077, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.01365661434829235, + "weighted_rkd_loss": 0.004477839567698538 + }, + { + "epoch": 7.567114093959732, + "flow_matching_loss": 0.015886571630835532, + "grad_norm": 0.14739073812961578, + "learning_rate": 7.840013258182471e-07, + "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": 0.0040415175259113315, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0015700169489718974, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.009653051989153028, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.015886571630835532, + "weighted_rkd_loss": 0.004826525994576514 + }, + { + "epoch": 7.583892617449664, + "flow_matching_loss": 0.015283238515257836, + "grad_norm": 0.18063873052597046, + "learning_rate": 7.526587435889809e-07, + "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": 0.004281807481311262, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0019427814171649516, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.010506396507844329, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.015283238515257836, + "weighted_rkd_loss": 0.005253198253922164 + }, + { + "epoch": 7.600671140939597, + "flow_matching_loss": 0.015953387320041656, + "grad_norm": 0.1318785548210144, + "learning_rate": 7.219393862212593e-07, + "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": 0.004065710003487766, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.001621024962514639, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.009752445016056299, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.015953387320041656, + "weighted_rkd_loss": 0.004876222508028149 + }, + { + "epoch": 7.617449664429531, + "flow_matching_loss": 0.015033760201185941, + "grad_norm": 0.18385183811187744, + "learning_rate": 6.918445974792819e-07, + "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": 0.004502101079560816, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0018878166214562953, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.010892018815502525, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.015033760201185941, + "weighted_rkd_loss": 0.005446009407751262 + }, + { + "epoch": 7.634228187919463, + "flow_matching_loss": 0.015050337370485067, + "grad_norm": 0.1594400852918625, + "learning_rate": 6.6237569380661e-07, + "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": 0.004070810438133776, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.00172933719586581, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.009870958141982555, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.015050337370485067, + "weighted_rkd_loss": 0.0049354790709912775 + }, + { + "epoch": 7.651006711409396, + "flow_matching_loss": 0.014827259257435799, + "grad_norm": 0.21137267351150513, + "learning_rate": 6.335339642685495e-07, + "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": 0.0043532455572858455, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.001815312390681356, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.010521803423762322, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.014827259257435799, + "weighted_rkd_loss": 0.005260901711881161 + }, + { + "epoch": 7.667785234899329, + "flow_matching_loss": 0.014157627988606692, + "grad_norm": 0.15432624518871307, + "learning_rate": 6.05320670495787e-07, + "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": 0.003650864795781672, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0015314153046347202, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.008833145024254918, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.014157627988606692, + "weighted_rkd_loss": 0.004416572512127459 + }, + { + "epoch": 7.684563758389261, + "flow_matching_loss": 0.013985906587913632, + "grad_norm": 0.12158872187137604, + "learning_rate": 5.777370466291954e-07, + "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": 0.003960758331231773, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0015997558715753258, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.009521272592246532, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.013985906587913632, + "weighted_rkd_loss": 0.004760636296123266 + }, + { + "epoch": 7.701342281879195, + "flow_matching_loss": 0.014527055993676185, + "grad_norm": 0.11905797570943832, + "learning_rate": 5.507842992658336e-07, + "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": 0.003300216724164784, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.001481162616983056, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.008081595972180366, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.014527055993676185, + "weighted_rkd_loss": 0.004040797986090183 + }, + { + "epoch": 7.718120805369128, + "flow_matching_loss": 0.01651501702144742, + "grad_norm": 0.144264355301857, + "learning_rate": 5.24463607406192e-07, + "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": 0.004112446680665016, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.001520359597634524, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.009745252970606088, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.01651501702144742, + "weighted_rkd_loss": 0.004872626485303044 + }, + { + "epoch": 7.73489932885906, + "flow_matching_loss": 0.013798393215984106, + "grad_norm": 0.18188291788101196, + "learning_rate": 4.987761224025944e-07, + "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": 0.004624183708801866, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0015590140363201499, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.010807381616905332, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.013798393215984106, + "weighted_rkd_loss": 0.005403690808452666 + }, + { + "epoch": 7.751677852348993, + "flow_matching_loss": 0.018524532951414586, + "grad_norm": 0.12207607179880142, + "learning_rate": 4.737229679088556e-07, + "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": 0.004238227335736156, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.001529688062146306, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.010006142687052489, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.018524532951414586, + "weighted_rkd_loss": 0.005003071343526244 + }, + { + "epoch": 7.768456375838926, + "flow_matching_loss": 0.01555617032572627, + "grad_norm": 0.2024114578962326, + "learning_rate": 4.493052398311115e-07, + "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": 0.00375926629640162, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.001470496621914208, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.008989029238000512, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.01555617032572627, + "weighted_rkd_loss": 0.004494514619000256 + }, + { + "epoch": 7.785234899328859, + "flow_matching_loss": 0.015568674076348543, + "grad_norm": 0.14492373168468475, + "learning_rate": 4.2552400627989043e-07, + "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": 0.0039431073004379865, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.001659181946888566, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.00954539654776454, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.015568674076348543, + "weighted_rkd_loss": 0.00477269827388227 + }, + { + "epoch": 7.802013422818792, + "flow_matching_loss": 0.01527152806520462, + "grad_norm": 0.1488606482744217, + "learning_rate": 4.023803075233923e-07, + "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": 0.0033758064033463596, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.001500186522025615, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.008251799317076803, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.01527152806520462, + "weighted_rkd_loss": 0.0041258996585384015 + }, + { + "epoch": 7.818791946308725, + "flow_matching_loss": 0.015966273937374355, + "grad_norm": 0.12519210577011108, + "learning_rate": 3.7987515594196997e-07, + "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": 0.004246301972307265, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0016014743596315384, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.010094078304246069, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.015966273937374355, + "weighted_rkd_loss": 0.005047039152123034 + }, + { + "epoch": 7.8355704697986575, + "flow_matching_loss": 0.015654311794787646, + "grad_norm": 0.19098609685897827, + "learning_rate": 3.580095359838648e-07, + "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": 0.0036975160473957658, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0014155545737594365, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.008810586575418711, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.015654311794787646, + "weighted_rkd_loss": 0.004405293287709356 + }, + { + "epoch": 7.85234899328859, + "flow_matching_loss": 0.015151944477111101, + "grad_norm": 0.131342351436615, + "learning_rate": 3.3678440412212576e-07, + "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": 0.00355694442987442, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0013498203014023602, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.008463709149509668, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.015151944477111101, + "weighted_rkd_loss": 0.004231854574754834 + }, + { + "epoch": 7.869127516778524, + "flow_matching_loss": 0.015845899749547244, + "grad_norm": 0.17855486273765564, + "learning_rate": 3.162006888127833e-07, + "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": 0.004481368698179722, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.001695212651975453, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.010657949931919574, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.015845899749547244, + "weighted_rkd_loss": 0.005328974965959787 + }, + { + "epoch": 7.885906040268456, + "flow_matching_loss": 0.017545492481440307, + "grad_norm": 0.1288793385028839, + "learning_rate": 2.9625929045422327e-07, + "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": 0.004457078292034566, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0017358578566927462, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.010650014318525792, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.017545492481440307, + "weighted_rkd_loss": 0.005325007159262896 + }, + { + "epoch": 7.902684563758389, + "flow_matching_loss": 0.01548297619447112, + "grad_norm": 0.15775883197784424, + "learning_rate": 2.7696108134781186e-07, + "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": 0.004006052413024009, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0015338835888542234, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.009545988403260708, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.01548297619447112, + "weighted_rkd_loss": 0.004772994201630354 + }, + { + "epoch": 7.919463087248322, + "flow_matching_loss": 0.017314035911113024, + "grad_norm": 0.17367108166217804, + "learning_rate": 2.583069056597376e-07, + "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": 0.00400707998778671, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0016400735126808285, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.009654233604669571, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.017314035911113024, + "weighted_rkd_loss": 0.004827116802334786 + }, + { + "epoch": 7.936241610738255, + "flow_matching_loss": 0.013726966083049774, + "grad_norm": 0.1574770212173462, + "learning_rate": 2.4029757938407116e-07, + "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": 0.004623283655382693, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0019199330708943307, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.011166500393301249, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.013726966083049774, + "weighted_rkd_loss": 0.005583250196650624 + }, + { + "epoch": 7.953020134228188, + "flow_matching_loss": 0.01642540153115988, + "grad_norm": 0.20747427642345428, + "learning_rate": 2.229338903070871e-07, + "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": 0.003777121054008603, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0015185043797828258, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.0090727464761585, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.01642540153115988, + "weighted_rkd_loss": 0.00453637323807925 + }, + { + "epoch": 7.969798657718121, + "flow_matching_loss": 0.01649800632148981, + "grad_norm": 0.23679505288600922, + "learning_rate": 2.0621659797279645e-07, + "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": 0.004173471452668309, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0016695614729542285, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.010016504442319274, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.01649800632148981, + "weighted_rkd_loss": 0.005008252221159637 + }, + { + "epoch": 7.9865771812080535, + "flow_matching_loss": 0.015903161372989416, + "grad_norm": 0.12486451864242554, + "learning_rate": 1.901464336497233e-07, + "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": 0.004392204666510224, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.001676009374205023, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01046041864901781, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.015903161372989416, + "weighted_rkd_loss": 0.005230209324508905 + }, + { + "epoch": 8.003355704697986, + "flow_matching_loss": 0.014692566078156232, + "grad_norm": 0.12237241864204407, + "learning_rate": 1.7472410029891527e-07, + "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": 0.0040526770986616615, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0014504189952276648, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.009555773250758648, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.014692566078156232, + "weighted_rkd_loss": 0.004777886625379324 + }, + { + "epoch": 8.020134228187919, + "flow_matching_loss": 0.014108181092888116, + "grad_norm": 0.10870523005723953, + "learning_rate": 1.5995027254319173e-07, + "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": 0.0040929106064140795, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.001453663141001016, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.009639484342187643, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.014108181092888116, + "weighted_rkd_loss": 0.0048197421710938215 + }, + { + "epoch": 8.036912751677852, + "flow_matching_loss": 0.0141829714179039, + "grad_norm": 0.12840935587882996, + "learning_rate": 1.4582559663764372e-07, + "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": 0.0034840189618989825, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0014725176617503167, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.008440555585548282, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.0141829714179039, + "weighted_rkd_loss": 0.004220277792774141 + }, + { + "epoch": 8.053691275167786, + "flow_matching_loss": 0.016363996639847757, + "grad_norm": 0.13803021609783173, + "learning_rate": 1.323506904413485e-07, + "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": 0.003999671293422579, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0015343882550951093, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.009533730894327163, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.016363996639847757, + "weighted_rkd_loss": 0.004766865447163582 + }, + { + "epoch": 8.070469798657719, + "flow_matching_loss": 0.014740627631545067, + "grad_norm": 0.15675628185272217, + "learning_rate": 1.195261433903594e-07, + "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": 0.0036474255612120032, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0014107772847637534, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.008705628290772437, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.014740627631545067, + "weighted_rkd_loss": 0.004352814145386219 + }, + { + "epoch": 8.087248322147651, + "flow_matching_loss": 0.015853640623390675, + "grad_norm": 0.21039828658103943, + "learning_rate": 1.0735251647190969e-07, + "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": 0.004786642431281507, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0017930870177224278, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01136637181043625, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.015853640623390675, + "weighted_rkd_loss": 0.005683185905218125 + }, + { + "epoch": 8.104026845637584, + "flow_matching_loss": 0.013481860840693116, + "grad_norm": 0.14539654552936554, + "learning_rate": 9.583034219987408e-08, + "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": 0.0032999050919897853, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0014778119686525315, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.008077622065320611, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.013481860840693116, + "weighted_rkd_loss": 0.004038811032660306 + }, + { + "epoch": 8.120805369127517, + "flow_matching_loss": 0.012944562174379826, + "grad_norm": 0.1711987853050232, + "learning_rate": 8.496012459148717e-08, + "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": 0.00421827076934278, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.00177860067342408, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.010215142229571938, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.012944562174379826, + "weighted_rkd_loss": 0.005107571114785969 + }, + { + "epoch": 8.13758389261745, + "flow_matching_loss": 0.016942327935248615, + "grad_norm": 0.16165891289710999, + "learning_rate": 7.474233914527618e-08, + "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": 0.0038894150173291566, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0016811634763143958, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.00945999352261424, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.016942327935248615, + "weighted_rkd_loss": 0.00472999676130712 + }, + { + "epoch": 8.154362416107382, + "flow_matching_loss": 0.016499594505876302, + "grad_norm": 0.20303112268447876, + "learning_rate": 6.517743282028254e-08, + "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": 0.003583537461236119, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0014558300492353738, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.008622905006632209, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.016499594505876302, + "weighted_rkd_loss": 0.004311452503316104 + }, + { + "epoch": 8.171140939597315, + "flow_matching_loss": 0.01599147580564022, + "grad_norm": 0.17316727340221405, + "learning_rate": 5.6265824016492604e-08, + "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": 0.003953036200255155, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0017714102636091412, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.009677482675760985, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.01599147580564022, + "weighted_rkd_loss": 0.004838741337880492 + }, + { + "epoch": 8.187919463087248, + "flow_matching_loss": 0.016761266998946665, + "grad_norm": 0.1150829866528511, + "learning_rate": 4.800790255654719e-08, + "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": 0.0036187476478517057, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0013934143644291908, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.008630909631028772, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.016761266998946665, + "weighted_rkd_loss": 0.004315454815514386 + }, + { + "epoch": 8.204697986577182, + "flow_matching_loss": 0.017060472443699837, + "grad_norm": 0.19914576411247253, + "learning_rate": 4.040402966867862e-08, + "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": 0.00388106107711792, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0015067332715261728, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.00926885548979044, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.017060472443699837, + "weighted_rkd_loss": 0.00463442774489522 + }, + { + "epoch": 8.221476510067115, + "flow_matching_loss": 0.01452061990275979, + "grad_norm": 0.17426694929599762, + "learning_rate": 3.3454537970925014e-08, + "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": 0.003739560558460653, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0014982009073719383, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.008977322094142436, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.01452061990275979, + "weighted_rkd_loss": 0.004488661047071218 + }, + { + "epoch": 8.238255033557047, + "flow_matching_loss": 0.01707121105864644, + "grad_norm": 0.15827062726020813, + "learning_rate": 2.715973145656192e-08, + "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": 0.0036373403389006855, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0014872469764668494, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.008761927671730518, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.01707121105864644, + "weighted_rkd_loss": 0.004380963835865259 + }, + { + "epoch": 8.25503355704698, + "flow_matching_loss": 0.013726630993187428, + "grad_norm": 0.13560862839221954, + "learning_rate": 2.151988548082129e-08, + "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": 0.0038332486758008598, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0015756478009279818, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.009242145158350468, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.013726630993187428, + "weighted_rkd_loss": 0.004621072579175234 + }, + { + "epoch": 8.271812080536913, + "flow_matching_loss": 0.015442668460309505, + "grad_norm": 0.20286814868450165, + "learning_rate": 1.6535246748837772e-08, + "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": 0.0048883697018027306, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.002030312130227685, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.011807051626965404, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.015442668460309505, + "weighted_rkd_loss": 0.005903525813482702 + }, + { + "epoch": 8.288590604026846, + "flow_matching_loss": 0.015910115744918584, + "grad_norm": 0.1755928248167038, + "learning_rate": 1.2206033304855702e-08, + "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": 0.004257333232089877, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0016399717191234232, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.010154638253152371, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.015910115744918584, + "weighted_rkd_loss": 0.005077319126576186 + }, + { + "epoch": 8.305369127516778, + "flow_matching_loss": 0.017014620639383794, + "grad_norm": 0.21290400624275208, + "learning_rate": 8.532434522695031e-09, + "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": 0.004705283232033253, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0018529294757172465, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.011263495916500688, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.017014620639383794, + "weighted_rkd_loss": 0.005631747958250344 + }, + { + "epoch": 8.322147651006711, + "flow_matching_loss": 0.015421977546066046, + "grad_norm": 0.139173686504364, + "learning_rate": 5.5146110974679805e-09, + "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": 0.003743798425421119, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0014029679470695556, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.008890564925968647, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.015421977546066046, + "weighted_rkd_loss": 0.004445282462984324 + }, + { + "epoch": 8.338926174496644, + "flow_matching_loss": 0.01395677267573774, + "grad_norm": 0.1474369466304779, + "learning_rate": 3.1526950385496467e-09, + "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": 0.004145475127734244, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0015348056796938181, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.009825755842030048, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.01395677267573774, + "weighted_rkd_loss": 0.004912877921015024 + }, + { + "epoch": 8.355704697986576, + "flow_matching_loss": 0.014628880564123392, + "grad_norm": 0.1921273171901703, + "learning_rate": 1.4467896637959666e-09, + "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": 0.004135441291145981, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.001468336395919323, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.00973921879194677, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.014628880564123392, + "weighted_rkd_loss": 0.004869609395973385 + }, + { + "epoch": 8.37248322147651, + "flow_matching_loss": 0.01582834655418992, + "grad_norm": 0.17792509496212006, + "learning_rate": 3.9696959503898733e-10, + "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": 0.004110996657982468, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.001513004262233153, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.009734997665509582, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.01582834655418992, + "weighted_rkd_loss": 0.004867498832754791 + }, + { + "epoch": 8.389261744966444, + "flow_matching_loss": 0.015212259162217379, + "grad_norm": 0.18235495686531067, + "learning_rate": 3.280754806156416e-12, + "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": 0.0036568080307915808, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.00138852980453521, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.00870214574970305, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.015212259162217379, + "weighted_rkd_loss": 0.004351072874851525 + } + ], + "logging_steps": 10, + "max_steps": 5000, + "num_input_tokens_seen": 0, + "num_train_epochs": 9, + "save_steps": 5001, + "stateful_callbacks": { + "TrainerControl": { + "args": { + "should_epoch_stop": false, + "should_evaluate": false, + "should_log": false, + "should_save": true, + "should_training_stop": true + }, + "attributes": {} + } + }, + "total_flos": 0.0, + "train_batch_size": 64, + "trial_name": null, + "trial_params": null +} diff --git a/FewShot/39tasks/5shot/FromRKD-Timewar_Processingline_v2/seed2/default/resolved_config.yaml b/FewShot/39tasks/5shot/FromRKD-Timewar_Processingline_v2/seed2/default/resolved_config.yaml new file mode 100644 index 0000000000000000000000000000000000000000..d87d6f68ba630019ad840c5c70ec165c170302ab --- /dev/null +++ b/FewShot/39tasks/5shot/FromRKD-Timewar_Processingline_v2/seed2/default/resolved_config.yaml @@ -0,0 +1,38 @@ +experiment_name: fewshot_39tasks_5shot_from_rkd_timewar_processingline_v2_seed2 +policy_type: groot_RKDv3 +base_model_path: ${HOME}/groot_robocasa/robocasa_v2/TimewarpVAE/RKDv3_DA_APHead_from_Atomic26Baseline60k/default/phase3 +output_root: ${HOME}/groot_robocasa/robocasa_v2/FewShot/39tasks/5shot/FromRKD-Timewar_Processingline_v2/seed2 +dataset: + dataset_soup: fewshot_heldout_atomic39_human_5shot +training: + num_gpus: 1 + batch_size: 64 + seed: 42 + filter_key_seed: 2 + manifest_path: ${HOME}/clvla/benchmarks/robocasa_v2/my_scripts/Fewshot/manifests/fewshot_heldout_atomic39/shot5_seed2.json + save_total_limit: 4 +model: + rkd_enabled: true + rkd_fm_loss_weight: 1.0 + rkd_loss_weight: 0.5 + rkd_loss_weight_schedule: null +phases: +- name: train_5shot_seed2_processingline_5k + max_steps: 5000 + save_steps: 0 + save_step_list: + - 1000 + - 3000 + - 5000 + trainable: + preset: processing_line_only + tune_llm: false + tune_visual: false + tune_projector: false + tune_diffusion_model: false + losses: + rkd_enabled: true + rkd_fm_loss_weight: 1.0 + rkd_loss_weight: 0.5 + rkd_loss_weight_schedule: null +resolved_sweep: {} diff --git a/FewShot/39tasks/5shot/FromRKD-Timewar_Processingline_v2/seed2/default/seed2.yaml b/FewShot/39tasks/5shot/FromRKD-Timewar_Processingline_v2/seed2/default/seed2.yaml new file mode 100644 index 0000000000000000000000000000000000000000..a69687a2bda8de0a321e13fa0577b5cae00ad92d --- /dev/null +++ b/FewShot/39tasks/5shot/FromRKD-Timewar_Processingline_v2/seed2/default/seed2.yaml @@ -0,0 +1,37 @@ +experiment_name: fewshot_39tasks_5shot_from_rkd_timewar_processingline_v2_seed2 +policy_type: groot_RKDv3 +base_model_path: ${HOME}/groot_robocasa/robocasa_v2/TimewarpVAE/RKDv3_DA_APHead_from_Atomic26Baseline60k/default/phase3 +output_root: ${HOME}/groot_robocasa/robocasa_v2/FewShot/39tasks/5shot/FromRKD-Timewar_Processingline_v2/seed2 +dataset: + dataset_soup: fewshot_heldout_atomic39_human_5shot +training: + num_gpus: 1 + batch_size: 64 + seed: 42 + filter_key_seed: 2 + manifest_path: ${HOME}/clvla/benchmarks/robocasa_v2/my_scripts/Fewshot/manifests/fewshot_heldout_atomic39/shot5_seed2.json + save_total_limit: 4 +model: + rkd_enabled: true + rkd_fm_loss_weight: 1.0 + rkd_loss_weight: 0.5 + rkd_loss_weight_schedule: null +phases: +- name: train_5shot_seed2_processingline_5k + max_steps: 5000 + save_steps: 0 + save_step_list: + - 1000 + - 3000 + - 5000 + trainable: + preset: processing_line_only + tune_llm: false + tune_visual: false + tune_projector: false + tune_diffusion_model: false + losses: + rkd_enabled: true + rkd_fm_loss_weight: 1.0 + rkd_loss_weight: 0.5 + rkd_loss_weight_schedule: null diff --git a/FewShot/39tasks/5shot/FromRKD-Timewar_Processingline_v2/seed2/default/train_5shot_seed2_processingline_5k/config.json b/FewShot/39tasks/5shot/FromRKD-Timewar_Processingline_v2/seed2/default/train_5shot_seed2_processingline_5k/config.json new file mode 100644 index 0000000000000000000000000000000000000000..4740d0455bef31c469ee5dab6d4a47d64df61665 --- /dev/null +++ b/FewShot/39tasks/5shot/FromRKD-Timewar_Processingline_v2/seed2/default/train_5shot_seed2_processingline_5k/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": 0.5, + "rkd_loss_weight_end": 0.0, + "rkd_loss_weight_schedule": null, + "rkd_loss_weight_start": 0.1, + "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/FewShot/39tasks/5shot/FromRKD-Timewar_Processingline_v2/seed2/default/train_5shot_seed2_processingline_5k/model.safetensors.index.json b/FewShot/39tasks/5shot/FromRKD-Timewar_Processingline_v2/seed2/default/train_5shot_seed2_processingline_5k/model.safetensors.index.json new file mode 100644 index 0000000000000000000000000000000000000000..b20baed71f4d16a411223b70d45d934c6cbbfe97 --- /dev/null +++ b/FewShot/39tasks/5shot/FromRKD-Timewar_Processingline_v2/seed2/default/train_5shot_seed2_processingline_5k/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/FewShot/39tasks/5shot/FromRKD-Timewar_Processingline_v2/seed2/default/train_5shot_seed2_processingline_5k/trainer_state.json b/FewShot/39tasks/5shot/FromRKD-Timewar_Processingline_v2/seed2/default/train_5shot_seed2_processingline_5k/trainer_state.json new file mode 100644 index 0000000000000000000000000000000000000000..f3d870b707a0cfc3851072f9c1387a323fb7054e --- /dev/null +++ b/FewShot/39tasks/5shot/FromRKD-Timewar_Processingline_v2/seed2/default/train_5shot_seed2_processingline_5k/trainer_state.json @@ -0,0 +1,23043 @@ +{ + "best_global_step": null, + "best_metric": null, + "best_model_checkpoint": null, + "epoch": 8.547008547008547, + "eval_steps": 500, + "global_step": 5000, + "is_hyper_param_search": false, + "is_local_process_zero": true, + "is_world_process_zero": true, + "log_history": [ + { + "epoch": 0.017094017094017096, + "flow_matching_loss": 0.20608119964599608, + "grad_norm": 1.4477208852767944, + "learning_rate": 1.08e-06, + "loss": 0.4085, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.1441098630428314, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.11659819930791855, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.4048179268836975, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.20608119964599608, + "weighted_rkd_loss": 0.20240896344184875 + }, + { + "epoch": 0.03418803418803419, + "flow_matching_loss": 0.22261241227388381, + "grad_norm": 1.1392409801483154, + "learning_rate": 2.28e-06, + "loss": 0.4131, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.13591156229376794, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.10920900553464889, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.38103213608264924, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.22261241227388381, + "weighted_rkd_loss": 0.19051606804132462 + }, + { + "epoch": 0.05128205128205128, + "flow_matching_loss": 0.20390308946371077, + "grad_norm": 0.9336247444152832, + "learning_rate": 3.48e-06, + "loss": 0.388, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.1328342378139496, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.1025419719517231, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.3682104468345642, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.20390308946371077, + "weighted_rkd_loss": 0.1841052234172821 + }, + { + "epoch": 0.06837606837606838, + "flow_matching_loss": 0.2105909526348114, + "grad_norm": 0.8749905824661255, + "learning_rate": 4.68e-06, + "loss": 0.3894, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.12913469299674035, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.09934469908475876, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.35761408507823944, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.2105909526348114, + "weighted_rkd_loss": 0.17880704253911972 + }, + { + "epoch": 0.08547008547008547, + "flow_matching_loss": 0.1939498320221901, + "grad_norm": 0.8234826326370239, + "learning_rate": 5.8800000000000005e-06, + "loss": 0.3703, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.1269106835126877, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.09883174449205398, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.3526531159877777, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.1939498320221901, + "weighted_rkd_loss": 0.17632655799388885 + }, + { + "epoch": 0.10256410256410256, + "flow_matching_loss": 0.15859663859009743, + "grad_norm": 0.7916533946990967, + "learning_rate": 7.08e-06, + "loss": 0.3244, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.12063853591680526, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.09041309282183647, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.3316901683807373, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.15859663859009743, + "weighted_rkd_loss": 0.16584508419036864 + }, + { + "epoch": 0.11965811965811966, + "flow_matching_loss": 0.15783419758081435, + "grad_norm": 0.8823524117469788, + "learning_rate": 8.28e-06, + "loss": 0.3089, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.11185461878776551, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0784363217651844, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.30214556455612185, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.15783419758081435, + "weighted_rkd_loss": 0.15107278227806092 + }, + { + "epoch": 0.13675213675213677, + "flow_matching_loss": 0.13894206359982492, + "grad_norm": 0.941081702709198, + "learning_rate": 9.48e-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": 0.1060655303299427, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0731036301702261, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.28523469567298887, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.13894206359982492, + "weighted_rkd_loss": 0.14261734783649443 + }, + { + "epoch": 0.15384615384615385, + "flow_matching_loss": 0.12293280363082885, + "grad_norm": 0.8303114771842957, + "learning_rate": 1.068e-05, + "loss": 0.258, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.10080260336399079, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.06849895156919956, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.270104156434536, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.12293280363082885, + "weighted_rkd_loss": 0.135052078217268 + }, + { + "epoch": 0.17094017094017094, + "flow_matching_loss": 0.1046357847750187, + "grad_norm": 0.7039380669593811, + "learning_rate": 1.1880000000000001e-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": 0.09367859661579132, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.06181581430137158, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.24917300790548325, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.1046357847750187, + "weighted_rkd_loss": 0.12458650395274162 + }, + { + "epoch": 0.18803418803418803, + "flow_matching_loss": 0.09870888739824295, + "grad_norm": 0.6019426584243774, + "learning_rate": 1.308e-05, + "loss": 0.2142, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.08723139390349388, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.05651620887219906, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.23097899854183196, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.09870888739824295, + "weighted_rkd_loss": 0.11548949927091598 + }, + { + "epoch": 0.20512820512820512, + "flow_matching_loss": 0.08241323307156563, + "grad_norm": 0.6689212918281555, + "learning_rate": 1.428e-05, + "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": 0.08734692260622978, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0545398335903883, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.2292336791753769, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.08241323307156563, + "weighted_rkd_loss": 0.11461683958768845 + }, + { + "epoch": 0.2222222222222222, + "flow_matching_loss": 0.0766731396317482, + "grad_norm": 0.578585684299469, + "learning_rate": 1.548e-05, + "loss": 0.1806, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.07932209149003029, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.049220692738890647, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.20786487609148024, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.0766731396317482, + "weighted_rkd_loss": 0.10393243804574012 + }, + { + "epoch": 0.23931623931623933, + "flow_matching_loss": 0.08649998903274536, + "grad_norm": 0.6135573983192444, + "learning_rate": 1.6680000000000003e-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": 0.07297333218157291, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0444162305444479, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.1903628960251808, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.08649998903274536, + "weighted_rkd_loss": 0.0951814480125904 + }, + { + "epoch": 0.2564102564102564, + "flow_matching_loss": 0.07530920952558517, + "grad_norm": 0.6404624581336975, + "learning_rate": 1.7879999999999998e-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": 0.06675466485321521, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.039724042639136314, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.17323337346315384, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.07530920952558517, + "weighted_rkd_loss": 0.08661668673157692 + }, + { + "epoch": 0.27350427350427353, + "flow_matching_loss": 0.07482291832566261, + "grad_norm": 0.614500880241394, + "learning_rate": 1.908e-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": 0.0673632200807333, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.040419120155274865, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.17514555901288986, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.07482291832566261, + "weighted_rkd_loss": 0.08757277950644493 + }, + { + "epoch": 0.2905982905982906, + "flow_matching_loss": 0.07167641781270503, + "grad_norm": 0.5213647484779358, + "learning_rate": 2.0280000000000002e-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": 0.0648277897387743, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.03784646335989237, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.16750204414129258, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.07167641781270503, + "weighted_rkd_loss": 0.08375102207064629 + }, + { + "epoch": 0.3076923076923077, + "flow_matching_loss": 0.06723086908459663, + "grad_norm": 0.5122416019439697, + "learning_rate": 2.148e-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": 0.05934748239815235, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.034743427857756616, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.15343839228153228, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.06723086908459663, + "weighted_rkd_loss": 0.07671919614076614 + }, + { + "epoch": 0.3247863247863248, + "flow_matching_loss": 0.06515696458518505, + "grad_norm": 0.5083960890769958, + "learning_rate": 2.268e-05, + "loss": 0.1374, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.055695648118853566, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.033016927726566794, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.14440822303295137, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.06515696458518505, + "weighted_rkd_loss": 0.07220411151647568 + }, + { + "epoch": 0.3418803418803419, + "flow_matching_loss": 0.06426175646483898, + "grad_norm": 0.5611753463745117, + "learning_rate": 2.3880000000000002e-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": 0.053841910883784296, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.03130413573235273, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.13898795917630197, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.06426175646483898, + "weighted_rkd_loss": 0.06949397958815098 + }, + { + "epoch": 0.358974358974359, + "flow_matching_loss": 0.061613532900810244, + "grad_norm": 0.4719548523426056, + "learning_rate": 2.508e-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": 0.048743893206119535, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.02855243906378746, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.1260402247309685, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.061613532900810244, + "weighted_rkd_loss": 0.06302011236548424 + }, + { + "epoch": 0.37606837606837606, + "flow_matching_loss": 0.058244922384619716, + "grad_norm": 0.489758163690567, + "learning_rate": 2.628e-05, + "loss": 0.1197, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.04813622012734413, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.026654975302517415, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.12292741760611534, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.058244922384619716, + "weighted_rkd_loss": 0.06146370880305767 + }, + { + "epoch": 0.39316239316239315, + "flow_matching_loss": 0.05867415480315685, + "grad_norm": 0.5043800473213196, + "learning_rate": 2.748e-05, + "loss": 0.1202, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.04817534200847149, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.02667076177895069, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.1230214461684227, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.05867415480315685, + "weighted_rkd_loss": 0.06151072308421135 + }, + { + "epoch": 0.41025641025641024, + "flow_matching_loss": 0.059197704493999484, + "grad_norm": 0.4292409420013428, + "learning_rate": 2.868e-05, + "loss": 0.1177, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.04579513482749462, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.025458005629479884, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.11704827472567558, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.059197704493999484, + "weighted_rkd_loss": 0.05852413736283779 + }, + { + "epoch": 0.42735042735042733, + "flow_matching_loss": 0.05531383752822876, + "grad_norm": 0.43816468119621277, + "learning_rate": 2.9880000000000002e-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": 0.046195313706994055, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.026640130765736104, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.1190307579934597, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.05531383752822876, + "weighted_rkd_loss": 0.05951537899672985 + }, + { + "epoch": 0.4444444444444444, + "flow_matching_loss": 0.05735364258289337, + "grad_norm": 0.4437594413757324, + "learning_rate": 2.9999734259635676e-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": 0.03679828383028507, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.020236621610820293, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.09383318945765495, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.05735364258289337, + "weighted_rkd_loss": 0.046916594728827477 + }, + { + "epoch": 0.46153846153846156, + "flow_matching_loss": 0.059609442576766015, + "grad_norm": 0.5054996013641357, + "learning_rate": 2.9998815663057245e-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": 0.041301059536635876, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.021624139230698346, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.10422625914216041, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.059609442576766015, + "weighted_rkd_loss": 0.052113129571080205 + }, + { + "epoch": 0.47863247863247865, + "flow_matching_loss": 0.05288516990840435, + "grad_norm": 0.48650580644607544, + "learning_rate": 2.999724096969228e-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": 0.04097450710833073, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.021585627086460592, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.10353464037179946, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.05288516990840435, + "weighted_rkd_loss": 0.05176732018589973 + }, + { + "epoch": 0.49572649572649574, + "flow_matching_loss": 0.0536977294832468, + "grad_norm": 0.42119359970092773, + "learning_rate": 2.9995010248422984e-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": 0.038251850940287116, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.01963985590264201, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.09614355638623237, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.0536977294832468, + "weighted_rkd_loss": 0.048071778193116185 + }, + { + "epoch": 0.5128205128205128, + "flow_matching_loss": 0.048235632851719855, + "grad_norm": 0.4279158115386963, + "learning_rate": 2.999212359682832e-05, + "loss": 0.1011, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.04124627523124218, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.023284840025007726, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.10577739253640175, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.048235632851719855, + "weighted_rkd_loss": 0.05288869626820088 + }, + { + "epoch": 0.5299145299145299, + "flow_matching_loss": 0.04915821589529514, + "grad_norm": 0.4583510756492615, + "learning_rate": 2.9988581141179787e-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": 0.03272451851516962, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.018078654445707797, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.08352769166231155, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.04915821589529514, + "weighted_rkd_loss": 0.04176384583115578 + }, + { + "epoch": 0.5470085470085471, + "flow_matching_loss": 0.0478827927261591, + "grad_norm": 0.4557395875453949, + "learning_rate": 2.998438303643587e-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": 0.03672756850719452, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.01931789694353938, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.09277303442358971, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.0478827927261591, + "weighted_rkd_loss": 0.046386517211794855 + }, + { + "epoch": 0.5641025641025641, + "flow_matching_loss": 0.048558805510401724, + "grad_norm": 0.4305928349494934, + "learning_rate": 2.997952946623527e-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": 0.03552434463053942, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.019004741962999106, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.09005343131721019, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.048558805510401724, + "weighted_rkd_loss": 0.045026715658605096 + }, + { + "epoch": 0.5811965811965812, + "flow_matching_loss": 0.05098775625228882, + "grad_norm": 0.40920335054397583, + "learning_rate": 2.9974020642888875e-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": 0.03249767124652862, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.01662399461492896, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.08161933720111847, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.05098775625228882, + "weighted_rkd_loss": 0.040809668600559235 + }, + { + "epoch": 0.5982905982905983, + "flow_matching_loss": 0.047236426174640654, + "grad_norm": 0.4020651578903198, + "learning_rate": 2.9967856807370458e-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": 0.032064920477569106, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.016578875295817853, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.08070871606469154, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.047236426174640654, + "weighted_rkd_loss": 0.04035435803234577 + }, + { + "epoch": 0.6153846153846154, + "flow_matching_loss": 0.049100768938660624, + "grad_norm": 0.410824179649353, + "learning_rate": 2.996103822930615e-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": 0.03171112034469843, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.015830404963344334, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.07925264574587346, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.049100768938660624, + "weighted_rkd_loss": 0.03962632287293673 + }, + { + "epoch": 0.6324786324786325, + "flow_matching_loss": 0.04570407420396805, + "grad_norm": 0.41649773716926575, + "learning_rate": 2.9953565206962653e-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": 0.03249792978167534, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.017130458448082208, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.08212631791830063, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.04570407420396805, + "weighted_rkd_loss": 0.04106315895915032 + }, + { + "epoch": 0.6495726495726496, + "flow_matching_loss": 0.04507583752274513, + "grad_norm": 0.41268017888069153, + "learning_rate": 2.9945438067234172e-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": 0.029481365904212, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.014191387593746186, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.07315411977469921, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.04507583752274513, + "weighted_rkd_loss": 0.036577059887349606 + }, + { + "epoch": 0.6666666666666666, + "flow_matching_loss": 0.04255103617906571, + "grad_norm": 0.38128671050071716, + "learning_rate": 2.9936657165628124e-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": 0.032352514937520026, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.016295951325446367, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.08100098073482513, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.04255103617906571, + "weighted_rkd_loss": 0.040500490367412566 + }, + { + "epoch": 0.6837606837606838, + "flow_matching_loss": 0.04447304848581553, + "grad_norm": 0.36115458607673645, + "learning_rate": 2.992722288624959e-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": 0.02881138101220131, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.01579931229352951, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.07342207506299019, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.04447304848581553, + "weighted_rkd_loss": 0.03671103753149509 + }, + { + "epoch": 0.7008547008547008, + "flow_matching_loss": 0.04228375982493162, + "grad_norm": 0.27871495485305786, + "learning_rate": 2.9917135641784515e-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": 0.028570003621280194, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.014606839325278998, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.07174684777855873, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.04228375982493162, + "weighted_rkd_loss": 0.03587342388927937 + }, + { + "epoch": 0.717948717948718, + "flow_matching_loss": 0.044394811242818834, + "grad_norm": 0.43788713216781616, + "learning_rate": 2.9906395873481655e-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": 0.0254867784678936, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.013939346186816692, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.06491290293633938, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.044394811242818834, + "weighted_rkd_loss": 0.03245645146816969 + }, + { + "epoch": 0.7350427350427351, + "flow_matching_loss": 0.04304563030600548, + "grad_norm": 0.32673171162605286, + "learning_rate": 2.989500405113326e-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": 0.026353750936686994, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.012691938411444425, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.06539944037795067, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.04304563030600548, + "weighted_rkd_loss": 0.03269972018897534 + }, + { + "epoch": 0.7521367521367521, + "flow_matching_loss": 0.04123639427125454, + "grad_norm": 0.33855384588241577, + "learning_rate": 2.9882960673054547e-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": 0.02597996760159731, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.012860519997775554, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.06482045613229274, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.04123639427125454, + "weighted_rkd_loss": 0.03241022806614637 + }, + { + "epoch": 0.7692307692307693, + "flow_matching_loss": 0.04132717214524746, + "grad_norm": 0.3718223571777344, + "learning_rate": 2.9870266266061886e-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": 0.026512700691819192, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.012095902115106583, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.06512130312621593, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.04132717214524746, + "weighted_rkd_loss": 0.032560651563107966 + }, + { + "epoch": 0.7863247863247863, + "flow_matching_loss": 0.042681030184030536, + "grad_norm": 0.3572709560394287, + "learning_rate": 2.9856921385449766e-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": 0.02780984025448561, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.013740665093064309, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.06936034522950649, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.042681030184030536, + "weighted_rkd_loss": 0.034680172614753244 + }, + { + "epoch": 0.8034188034188035, + "flow_matching_loss": 0.042522648349404335, + "grad_norm": 0.3548585772514343, + "learning_rate": 2.9842926614966502e-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": 0.023204243928194045, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.011549634858965874, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.057958122715353964, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.042522648349404335, + "weighted_rkd_loss": 0.028979061357676982 + }, + { + "epoch": 0.8205128205128205, + "flow_matching_loss": 0.03533885534852743, + "grad_norm": 0.3358217775821686, + "learning_rate": 2.982828256678868e-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": 0.024509145691990853, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.01339802285656333, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.06241631470620632, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.03533885534852743, + "weighted_rkd_loss": 0.03120815735310316 + }, + { + "epoch": 0.8376068376068376, + "flow_matching_loss": 0.0407237846404314, + "grad_norm": 0.40146124362945557, + "learning_rate": 2.9812989881494412e-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": 0.027627827040851115, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.013586688507348298, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.06884234212338924, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.0407237846404314, + "weighted_rkd_loss": 0.03442117106169462 + }, + { + "epoch": 0.8547008547008547, + "flow_matching_loss": 0.042021133750677106, + "grad_norm": 0.2975722849369049, + "learning_rate": 2.9797049228035296e-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": 0.02429118975996971, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.01214608708396554, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.060728466510772704, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.042021133750677106, + "weighted_rkd_loss": 0.030364233255386352 + }, + { + "epoch": 0.8717948717948718, + "flow_matching_loss": 0.03959287479519844, + "grad_norm": 0.37908297777175903, + "learning_rate": 2.978046130370715e-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": 0.02314175870269537, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.011852209083735942, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.0581357266753912, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.03959287479519844, + "weighted_rkd_loss": 0.0290678633376956 + }, + { + "epoch": 0.8888888888888888, + "flow_matching_loss": 0.038183963485062124, + "grad_norm": 0.35484832525253296, + "learning_rate": 2.9763226834119535e-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": 0.022654735110700132, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.01152317225933075, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.05683264210820198, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.038183963485062124, + "weighted_rkd_loss": 0.02841632105410099 + }, + { + "epoch": 0.905982905982906, + "flow_matching_loss": 0.03838554341346025, + "grad_norm": 0.3829350471496582, + "learning_rate": 2.9745346573163963e-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": 0.02331932317465544, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.012097928486764432, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.058736575022339824, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.03838554341346025, + "weighted_rkd_loss": 0.029368287511169912 + }, + { + "epoch": 0.9230769230769231, + "flow_matching_loss": 0.0373695008456707, + "grad_norm": 0.39486923813819885, + "learning_rate": 2.9726821302980987e-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": 0.02153742704540491, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.011604637745767832, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.054679491370916364, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.0373695008456707, + "weighted_rkd_loss": 0.027339745685458182 + }, + { + "epoch": 0.9401709401709402, + "flow_matching_loss": 0.03420814126729965, + "grad_norm": 0.30264589190483093, + "learning_rate": 2.9707651833925935e-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": 0.020150486007332803, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.009803551714867353, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.050104524567723276, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.03420814126729965, + "weighted_rkd_loss": 0.025052262283861638 + }, + { + "epoch": 0.9572649572649573, + "flow_matching_loss": 0.03828359097242355, + "grad_norm": 0.27153193950653076, + "learning_rate": 2.968783900453348e-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": 0.02007452305406332, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.009622522396966815, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.049771569296717645, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.03828359097242355, + "weighted_rkd_loss": 0.024885784648358823 + }, + { + "epoch": 0.9743589743589743, + "flow_matching_loss": 0.03746143244206905, + "grad_norm": 0.28983309864997864, + "learning_rate": 2.9667383681480968e-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": 0.02210656674578786, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.010438397945836186, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.05465153120458126, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.03746143244206905, + "weighted_rkd_loss": 0.02732576560229063 + }, + { + "epoch": 0.9914529914529915, + "flow_matching_loss": 0.035928660072386265, + "grad_norm": 0.30345749855041504, + "learning_rate": 2.9646286759550493e-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": 0.021595945302397012, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.010564278997480869, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.05375616922974587, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.035928660072386265, + "weighted_rkd_loss": 0.026878084614872934 + }, + { + "epoch": 1.0085470085470085, + "flow_matching_loss": 0.03552918881177902, + "grad_norm": 0.24723964929580688, + "learning_rate": 2.9624549161589764e-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": 0.018401246704161166, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.008261607168242336, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.0450641006231308, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.03552918881177902, + "weighted_rkd_loss": 0.0225320503115654 + }, + { + "epoch": 1.0256410256410255, + "flow_matching_loss": 0.03407776262611151, + "grad_norm": 0.30071568489074707, + "learning_rate": 2.9602171838471745e-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": 0.019089525938034056, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.009438153170049191, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.04761720485985279, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.03407776262611151, + "weighted_rkd_loss": 0.023808602429926395 + }, + { + "epoch": 1.0427350427350428, + "flow_matching_loss": 0.03366336897015572, + "grad_norm": 0.302155464887619, + "learning_rate": 2.957915576905303e-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": 0.02271198071539402, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.011890822742134332, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.057314783707261084, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.03366336897015572, + "weighted_rkd_loss": 0.028657391853630542 + }, + { + "epoch": 1.0598290598290598, + "flow_matching_loss": 0.03334424737840891, + "grad_norm": 0.31046783924102783, + "learning_rate": 2.9555501960131058e-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": 0.019524640496820213, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.009662961680442094, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.04871224202215672, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.03334424737840891, + "weighted_rkd_loss": 0.02435612101107836 + }, + { + "epoch": 1.0769230769230769, + "flow_matching_loss": 0.03312153853476048, + "grad_norm": 0.28929588198661804, + "learning_rate": 2.9531211446400068e-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": 0.017922693397849798, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.00835406300611794, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.044199449568986894, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.03312153853476048, + "weighted_rkd_loss": 0.022099724784493447 + }, + { + "epoch": 1.0940170940170941, + "flow_matching_loss": 0.031021049618721007, + "grad_norm": 0.29626306891441345, + "learning_rate": 2.9506285290405813e-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": 0.018444733694195747, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.008001519087702035, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.0448909867554903, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.031021049618721007, + "weighted_rkd_loss": 0.02244549337774515 + }, + { + "epoch": 1.1111111111111112, + "flow_matching_loss": 0.032624239847064015, + "grad_norm": 0.2606738209724426, + "learning_rate": 2.948072458249911e-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": 0.01954005965963006, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.010045465733855962, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.04912558533251286, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.032624239847064015, + "weighted_rkd_loss": 0.02456279266625643 + }, + { + "epoch": 1.1282051282051282, + "flow_matching_loss": 0.0357012577354908, + "grad_norm": 0.35832640528678894, + "learning_rate": 2.9454530440788125e-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": 0.01891506966203451, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.008951532281935215, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.04678167216479778, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.0357012577354908, + "weighted_rkd_loss": 0.02339083608239889 + }, + { + "epoch": 1.1452991452991452, + "flow_matching_loss": 0.035934435203671455, + "grad_norm": 0.3307911455631256, + "learning_rate": 2.9427704011089474e-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": 0.018776980880647897, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.008833648543804884, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.046387610211968425, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.035934435203671455, + "weighted_rkd_loss": 0.023193805105984212 + }, + { + "epoch": 1.1623931623931625, + "flow_matching_loss": 0.035431265085935595, + "grad_norm": 0.32725879549980164, + "learning_rate": 2.94002464668781e-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": 0.018233825359493493, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.008717586379498243, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.045185237750411035, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.035431265085935595, + "weighted_rkd_loss": 0.022592618875205518 + }, + { + "epoch": 1.1794871794871795, + "flow_matching_loss": 0.03349699769169092, + "grad_norm": 0.3060956299304962, + "learning_rate": 2.9372159009235936e-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": 0.017201956920325757, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.008407466765493154, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.042811380326747896, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.03349699769169092, + "weighted_rkd_loss": 0.021405690163373948 + }, + { + "epoch": 1.1965811965811965, + "flow_matching_loss": 0.037000064924359324, + "grad_norm": 0.2392890304327011, + "learning_rate": 2.9343442866799365e-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": 0.01774967508390546, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.008203081460669637, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.04370243176817894, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.037000064924359324, + "weighted_rkd_loss": 0.02185121588408947 + }, + { + "epoch": 1.2136752136752136, + "flow_matching_loss": 0.0321138696745038, + "grad_norm": 0.2744523286819458, + "learning_rate": 2.9314099295705495e-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": 0.01787513755261898, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.008945518126711249, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.04469579309225082, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.0321138696745038, + "weighted_rkd_loss": 0.02234789654612541 + }, + { + "epoch": 1.2307692307692308, + "flow_matching_loss": 0.03291467633098364, + "grad_norm": 0.27376607060432434, + "learning_rate": 2.928412957953718e-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": 0.0181625971570611, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.008992884447798133, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.04531807880848646, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.03291467633098364, + "weighted_rkd_loss": 0.02265903940424323 + }, + { + "epoch": 1.2478632478632479, + "flow_matching_loss": 0.031597768515348436, + "grad_norm": 0.29815053939819336, + "learning_rate": 2.925353502926689e-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": 0.015860845521092414, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.007615697477012873, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.03933738861232996, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.031597768515348436, + "weighted_rkd_loss": 0.01966869430616498 + }, + { + "epoch": 1.264957264957265, + "flow_matching_loss": 0.03209205660969019, + "grad_norm": 0.29948389530181885, + "learning_rate": 2.922231698319937e-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": 0.017416194640100004, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.008499196963384748, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.04333158619701862, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.03209205660969019, + "weighted_rkd_loss": 0.02166579309850931 + }, + { + "epoch": 1.282051282051282, + "flow_matching_loss": 0.02969713993370533, + "grad_norm": 0.2717636227607727, + "learning_rate": 2.9190476806913088e-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": 0.0162487150169909, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.007376636797562242, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.039874066784977916, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.02969713993370533, + "weighted_rkd_loss": 0.019937033392488958 + }, + { + "epoch": 1.2991452991452992, + "flow_matching_loss": 0.031202525086700916, + "grad_norm": 0.28742673993110657, + "learning_rate": 2.9158015893200504e-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": 0.017281018663197756, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.007276421086862684, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.04183845836669207, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.031202525086700916, + "weighted_rkd_loss": 0.020919229183346034 + }, + { + "epoch": 1.3162393162393162, + "flow_matching_loss": 0.03131227232515812, + "grad_norm": 0.28647977113723755, + "learning_rate": 2.9124935662007137e-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": 0.01633641654625535, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.007625138806179166, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.040297971293330195, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.03131227232515812, + "weighted_rkd_loss": 0.020148985646665098 + }, + { + "epoch": 1.3333333333333333, + "flow_matching_loss": 0.031031945906579494, + "grad_norm": 0.3045690655708313, + "learning_rate": 2.9091237560369467e-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": 0.01676619527861476, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.007660567946732044, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.041192957758903505, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.031031945906579494, + "weighted_rkd_loss": 0.020596478879451752 + }, + { + "epoch": 1.3504273504273505, + "flow_matching_loss": 0.03271263670176268, + "grad_norm": 0.2736513912677765, + "learning_rate": 2.905692306235163e-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": 0.01463465103879571, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.00730584766715765, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.03657514955848455, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.03271263670176268, + "weighted_rkd_loss": 0.018287574779242276 + }, + { + "epoch": 1.3675213675213675, + "flow_matching_loss": 0.030691403523087503, + "grad_norm": 0.28391724824905396, + "learning_rate": 2.9021993668980925e-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": 0.016008828580379487, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.008079428132623434, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.040097085386514665, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.030691403523087503, + "weighted_rkd_loss": 0.020048542693257333 + }, + { + "epoch": 1.3846153846153846, + "flow_matching_loss": 0.034423727355897424, + "grad_norm": 0.27780401706695557, + "learning_rate": 2.8986450908182185e-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": 0.016429053619503975, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.007567548984661698, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.04042565673589706, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.034423727355897424, + "weighted_rkd_loss": 0.02021282836794853 + }, + { + "epoch": 1.4017094017094016, + "flow_matching_loss": 0.03095700442790985, + "grad_norm": 0.22609981894493103, + "learning_rate": 2.8950296334710902e-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": 0.01725160414353013, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.007418723125010729, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.041921931132674214, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.03095700442790985, + "weighted_rkd_loss": 0.020960965566337107 + }, + { + "epoch": 1.4188034188034189, + "flow_matching_loss": 0.03410516288131475, + "grad_norm": 0.2484390139579773, + "learning_rate": 2.8913531530085248e-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": 0.015002989396452903, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.007663191389292478, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.037669170275330545, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.03410516288131475, + "weighted_rkd_loss": 0.018834585137665272 + }, + { + "epoch": 1.435897435897436, + "flow_matching_loss": 0.030957960151135923, + "grad_norm": 0.2720315754413605, + "learning_rate": 2.887615810251687e-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": 0.016805268358439207, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.007622529938817024, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.04123306684195995, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.030957960151135923, + "weighted_rkd_loss": 0.020616533420979976 + }, + { + "epoch": 1.452991452991453, + "flow_matching_loss": 0.030326111428439616, + "grad_norm": 0.2526715099811554, + "learning_rate": 2.8838177686840577e-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": 0.014052478969097138, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.006443384615704417, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.03454834185540676, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.030326111428439616, + "weighted_rkd_loss": 0.01727417092770338 + }, + { + "epoch": 1.4700854700854702, + "flow_matching_loss": 0.03168223239481449, + "grad_norm": 0.3726175129413605, + "learning_rate": 2.8799591944442776e-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": 0.01616350244730711, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.007998800138011575, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.04032580554485321, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.03168223239481449, + "weighted_rkd_loss": 0.020162902772426605 + }, + { + "epoch": 1.4871794871794872, + "flow_matching_loss": 0.030286968126893042, + "grad_norm": 0.24847525358200073, + "learning_rate": 2.8760402563188848e-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": 0.014799616392701864, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.007003285316750407, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.03660251796245575, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.030286968126893042, + "weighted_rkd_loss": 0.018301258981227874 + }, + { + "epoch": 1.5042735042735043, + "flow_matching_loss": 0.029968923516571523, + "grad_norm": 0.31482070684432983, + "learning_rate": 2.8720611257349285e-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": 0.01352565586566925, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.006156855309382081, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.03320816699415445, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.029968923516571523, + "weighted_rkd_loss": 0.016604083497077225 + }, + { + "epoch": 1.5213675213675213, + "flow_matching_loss": 0.03208826966583729, + "grad_norm": 0.26682037115097046, + "learning_rate": 2.868021976752471e-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": 0.012783072888851166, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.006290466827340424, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.031856612861156465, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.03208826966583729, + "weighted_rkd_loss": 0.015928306430578233 + }, + { + "epoch": 1.5384615384615383, + "flow_matching_loss": 0.029574750922620296, + "grad_norm": 0.2852403223514557, + "learning_rate": 2.8639229860569738e-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": 0.01526008890941739, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0071847605984658005, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.03770493865013123, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.029574750922620296, + "weighted_rkd_loss": 0.018852469325065613 + }, + { + "epoch": 1.5555555555555556, + "flow_matching_loss": 0.028749799728393553, + "grad_norm": 0.24580393731594086, + "learning_rate": 2.8597643329515688e-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": 0.012961071636527776, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.006020331801846623, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.031942475400865075, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.028749799728393553, + "weighted_rkd_loss": 0.015971237700432538 + }, + { + "epoch": 1.5726495726495726, + "flow_matching_loss": 0.032204396836459634, + "grad_norm": 0.20449334383010864, + "learning_rate": 2.855546199349214e-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": 0.014906360022723674, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.006444963859394193, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.0362576836720109, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.032204396836459634, + "weighted_rkd_loss": 0.01812884183600545 + }, + { + "epoch": 1.5897435897435899, + "flow_matching_loss": 0.029012378305196762, + "grad_norm": 0.2885199785232544, + "learning_rate": 2.85126876976474e-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": 0.015990046598017216, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.007635087100788951, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.03961518015712499, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.029012378305196762, + "weighted_rkd_loss": 0.019807590078562497 + }, + { + "epoch": 1.606837606837607, + "flow_matching_loss": 0.0270179970189929, + "grad_norm": 0.2731475830078125, + "learning_rate": 2.8469322313067726e-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": 0.014710118900984526, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.006910082930698991, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.03633032031357288, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.0270179970189929, + "weighted_rkd_loss": 0.01816516015678644 + }, + { + "epoch": 1.623931623931624, + "flow_matching_loss": 0.03037530407309532, + "grad_norm": 0.19273969531059265, + "learning_rate": 2.8425367736695522e-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": 0.014435206167399883, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.00685276158619672, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.03572317436337471, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.03037530407309532, + "weighted_rkd_loss": 0.017861587181687354 + }, + { + "epoch": 1.641025641025641, + "flow_matching_loss": 0.027571728639304637, + "grad_norm": 0.25559166073799133, + "learning_rate": 2.838082589124636e-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": 0.012530108634382486, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.006172595894895494, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.031232813000679018, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.027571728639304637, + "weighted_rkd_loss": 0.015616406500339509 + }, + { + "epoch": 1.658119658119658, + "flow_matching_loss": 0.026423130929470063, + "grad_norm": 0.2502352297306061, + "learning_rate": 2.8335698725124843e-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": 0.013723570108413696, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.00633350177668035, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.0337806424126029, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.026423130929470063, + "weighted_rkd_loss": 0.01689032120630145 + }, + { + "epoch": 1.6752136752136753, + "flow_matching_loss": 0.026968703232705594, + "grad_norm": 0.29313844442367554, + "learning_rate": 2.8289988212339422e-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": 0.015268613491207362, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.006759832962416112, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.0372970599681139, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.026968703232705594, + "weighted_rkd_loss": 0.01864852998405695 + }, + { + "epoch": 1.6923076923076923, + "flow_matching_loss": 0.025319616310298443, + "grad_norm": 0.2522859573364258, + "learning_rate": 2.8243696352415997e-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": 0.014617102593183518, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.006836280273273587, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.03607048541307449, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.025319616310298443, + "weighted_rkd_loss": 0.018035242706537245 + }, + { + "epoch": 1.7094017094017095, + "flow_matching_loss": 0.030002601444721222, + "grad_norm": 0.24149076640605927, + "learning_rate": 2.8196825170310495e-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": 0.01380759608000517, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.006553543242625892, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.03416873514652252, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.030002601444721222, + "weighted_rkd_loss": 0.01708436757326126 + }, + { + "epoch": 1.7264957264957266, + "flow_matching_loss": 0.02807784080505371, + "grad_norm": 0.25452151894569397, + "learning_rate": 2.8149376716320256e-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": 0.014782947115600109, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.00671710423193872, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.03628299850970507, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.02807784080505371, + "weighted_rkd_loss": 0.018141499254852533 + }, + { + "epoch": 1.7435897435897436, + "flow_matching_loss": 0.02710084058344364, + "grad_norm": 0.33143165707588196, + "learning_rate": 2.8101353065994373e-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": 0.014535702299326658, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.006557283597066999, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.03562868814915419, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.02710084058344364, + "weighted_rkd_loss": 0.017814344074577094 + }, + { + "epoch": 1.7606837606837606, + "flow_matching_loss": 0.02593045849353075, + "grad_norm": 0.3127861022949219, + "learning_rate": 2.8052756320042885e-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": 0.01566762672737241, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.007534478232264519, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.03886973131448031, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.02593045849353075, + "weighted_rkd_loss": 0.019434865657240154 + }, + { + "epoch": 1.7777777777777777, + "flow_matching_loss": 0.0273212568834424, + "grad_norm": 0.2180718183517456, + "learning_rate": 2.8003588604244893e-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": 0.01422199010848999, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.00723591735586524, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.035679897479712966, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.0273212568834424, + "weighted_rkd_loss": 0.017839948739856483 + }, + { + "epoch": 1.7948717948717947, + "flow_matching_loss": 0.027072438225150108, + "grad_norm": 0.2212599217891693, + "learning_rate": 2.7953852069355565e-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": 0.013318956829607487, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0059395078103989364, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.03257742114365101, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.027072438225150108, + "weighted_rkd_loss": 0.016288710571825505 + }, + { + "epoch": 1.811965811965812, + "flow_matching_loss": 0.026676180213689803, + "grad_norm": 0.2384062260389328, + "learning_rate": 2.790354889101206e-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": 0.011322647333145142, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.005065834778361022, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.027711129188537596, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.026676180213689803, + "weighted_rkd_loss": 0.013855564594268798 + }, + { + "epoch": 1.8290598290598292, + "flow_matching_loss": 0.028350574523210527, + "grad_norm": 0.26893094182014465, + "learning_rate": 2.7852681269638366e-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": 0.01224717302247882, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.005544273066334426, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.030038618855178355, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.028350574523210527, + "weighted_rkd_loss": 0.015019309427589177 + }, + { + "epoch": 1.8461538461538463, + "flow_matching_loss": 0.023936538957059383, + "grad_norm": 0.32209914922714233, + "learning_rate": 2.780125143034902e-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": 0.012277705408632756, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.00516571004409343, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.02972112111747265, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.023936538957059383, + "weighted_rkd_loss": 0.014860560558736325 + }, + { + "epoch": 1.8632478632478633, + "flow_matching_loss": 0.02863725144416094, + "grad_norm": 0.22413775324821472, + "learning_rate": 2.7749261622851812e-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": 0.012101378757506609, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.005476271221414209, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.029679028689861296, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.02863725144416094, + "weighted_rkd_loss": 0.014839514344930648 + }, + { + "epoch": 1.8803418803418803, + "flow_matching_loss": 0.027668011374771595, + "grad_norm": 0.26382219791412354, + "learning_rate": 2.7696714121349335e-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": 0.013915220554918051, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.005969888484105468, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.03380032908171415, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.027668011374771595, + "weighted_rkd_loss": 0.016900164540857077 + }, + { + "epoch": 1.8974358974358974, + "flow_matching_loss": 0.02955530472099781, + "grad_norm": 0.27939197421073914, + "learning_rate": 2.764361122443954e-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": 0.01350008575245738, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.00708389279898256, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.034084064327180384, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.02955530472099781, + "weighted_rkd_loss": 0.017042032163590192 + }, + { + "epoch": 1.9145299145299144, + "flow_matching_loss": 0.02437609266489744, + "grad_norm": 0.2628627419471741, + "learning_rate": 2.758995525501517e-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": 0.01297243358567357, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.005478202877566218, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.03142307046800852, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.02437609266489744, + "weighted_rkd_loss": 0.01571153523400426 + }, + { + "epoch": 1.9316239316239316, + "flow_matching_loss": 0.028911850042641164, + "grad_norm": 0.23939956724643707, + "learning_rate": 2.753574856016215e-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": 0.012397017190232873, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.005886134272441268, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.03068016842007637, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.028911850042641164, + "weighted_rkd_loss": 0.015340084210038185 + }, + { + "epoch": 1.9487179487179487, + "flow_matching_loss": 0.02508879117667675, + "grad_norm": 0.20714369416236877, + "learning_rate": 2.7480993511056913e-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": 0.01294971164315939, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0063501600408926604, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.03224958349019289, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.02508879117667675, + "weighted_rkd_loss": 0.016124791745096446 + }, + { + "epoch": 1.965811965811966, + "flow_matching_loss": 0.024527727998793124, + "grad_norm": 0.19642457365989685, + "learning_rate": 2.7425692502862684e-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": 0.011007771594449878, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0052364618517458435, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.027252004854381083, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.024527727998793124, + "weighted_rkd_loss": 0.013626002427190542 + }, + { + "epoch": 1.982905982905983, + "flow_matching_loss": 0.02737492471933365, + "grad_norm": 0.23660895228385925, + "learning_rate": 2.7369847954624728e-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": 0.0125096146017313, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0061733081936836244, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.031192537397146225, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.02737492471933365, + "weighted_rkd_loss": 0.015596268698573112 + }, + { + "epoch": 2.0, + "flow_matching_loss": 0.026675242371857165, + "grad_norm": 0.25652265548706055, + "learning_rate": 2.731346230916448e-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": 0.013060001470148564, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.006334699317812919, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.03245470225811005, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.026675242371857165, + "weighted_rkd_loss": 0.016227351129055025 + }, + { + "epoch": 2.017094017094017, + "flow_matching_loss": 0.025632075406610964, + "grad_norm": 0.22836749255657196, + "learning_rate": 2.7256538032972746e-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": 0.012878391984850168, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.005053568026050925, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.030810352042317392, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.025632075406610964, + "weighted_rkd_loss": 0.015405176021158696 + }, + { + "epoch": 2.034188034188034, + "flow_matching_loss": 0.027616259641945362, + "grad_norm": 0.22967053949832916, + "learning_rate": 2.7199077616101773e-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": 0.012935011554509402, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.005700286431238055, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.03157030940055847, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.027616259641945362, + "weighted_rkd_loss": 0.015785154700279237 + }, + { + "epoch": 2.051282051282051, + "flow_matching_loss": 0.02580515146255493, + "grad_norm": 0.2185813933610916, + "learning_rate": 2.714108357205634e-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": 0.013044429570436477, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.005809265840798617, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.03189812451601028, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.02580515146255493, + "weighted_rkd_loss": 0.01594906225800514 + }, + { + "epoch": 2.0683760683760686, + "flow_matching_loss": 0.02779768891632557, + "grad_norm": 0.21726644039154053, + "learning_rate": 2.7082558437683803e-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": 0.012998757604509591, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.005653172452002764, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.03165068775415421, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.02779768891632557, + "weighted_rkd_loss": 0.015825343877077104 + }, + { + "epoch": 2.0854700854700856, + "flow_matching_loss": 0.0263330465182662, + "grad_norm": 0.22308462858200073, + "learning_rate": 2.702350477306315e-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": 0.011554498877376317, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.005302437697537243, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.028411435894668102, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.0263330465182662, + "weighted_rkd_loss": 0.014205717947334051 + }, + { + "epoch": 2.1025641025641026, + "flow_matching_loss": 0.02595714498311281, + "grad_norm": 0.22663040459156036, + "learning_rate": 2.6963925161392963e-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": 0.011306988494470715, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.004992392286658287, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.027606369741261006, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.02595714498311281, + "weighted_rkd_loss": 0.013803184870630503 + }, + { + "epoch": 2.1196581196581197, + "flow_matching_loss": 0.024864679016172887, + "grad_norm": 0.3029778301715851, + "learning_rate": 2.6903822208878476e-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": 0.01236183405853808, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.00573265366256237, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.030456322245299816, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.024864679016172887, + "weighted_rkd_loss": 0.015228161122649908 + }, + { + "epoch": 2.1367521367521367, + "flow_matching_loss": 0.02325292509049177, + "grad_norm": 0.2267565280199051, + "learning_rate": 2.6843198544617535e-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": 0.011847892682999373, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.00517009780742228, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.028865883499383925, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.02325292509049177, + "weighted_rkd_loss": 0.014432941749691963 + }, + { + "epoch": 2.1538461538461537, + "flow_matching_loss": 0.02737160213291645, + "grad_norm": 0.20818153023719788, + "learning_rate": 2.6782056820485612e-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": 0.012656896701082587, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.00568916043266654, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.031002953462302685, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.02737160213291645, + "weighted_rkd_loss": 0.015501476731151343 + }, + { + "epoch": 2.1709401709401708, + "flow_matching_loss": 0.026486062444746494, + "grad_norm": 0.2249724268913269, + "learning_rate": 2.672039971101979e-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": 0.011717427894473076, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.005194938997738064, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.028629794716835022, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.026486062444746494, + "weighted_rkd_loss": 0.014314897358417511 + }, + { + "epoch": 2.1880341880341883, + "flow_matching_loss": 0.025420571491122246, + "grad_norm": 0.27121660113334656, + "learning_rate": 2.665822991330178e-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": 0.012049693893641234, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.005436556437052787, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.029535944014787673, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.025420571491122246, + "weighted_rkd_loss": 0.014767972007393836 + }, + { + "epoch": 2.2051282051282053, + "flow_matching_loss": 0.025054209306836127, + "grad_norm": 0.3212575316429138, + "learning_rate": 2.6595550146839934e-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": 0.011498598288744688, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0056119068060070275, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.02860910352319479, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.025054209306836127, + "weighted_rkd_loss": 0.014304551761597395 + }, + { + "epoch": 2.2222222222222223, + "flow_matching_loss": 0.02525439951568842, + "grad_norm": 0.26853254437446594, + "learning_rate": 2.6532363153450287e-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": 0.012216195557266473, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.005422738334164024, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.029855129588395358, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.02525439951568842, + "weighted_rkd_loss": 0.014927564794197679 + }, + { + "epoch": 2.2393162393162394, + "flow_matching_loss": 0.021937138680368663, + "grad_norm": 0.1946357786655426, + "learning_rate": 2.6468671697136634e-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": 0.01110919527709484, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.005047859624028206, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.02726624980568886, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.021937138680368663, + "weighted_rkd_loss": 0.01363312490284443 + }, + { + "epoch": 2.2564102564102564, + "flow_matching_loss": 0.02496455330401659, + "grad_norm": 0.22266526520252228, + "learning_rate": 2.640447856396961e-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": 0.011567748803645372, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0050113315926864745, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.02814682889729738, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.02496455330401659, + "weighted_rkd_loss": 0.01407341444864869 + }, + { + "epoch": 2.2735042735042734, + "flow_matching_loss": 0.02324515637010336, + "grad_norm": 0.22817914187908173, + "learning_rate": 2.6339786561964805e-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": 0.012382819782942533, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.005309685179963708, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.03007532488554716, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.02324515637010336, + "weighted_rkd_loss": 0.01503766244277358 + }, + { + "epoch": 2.2905982905982905, + "flow_matching_loss": 0.023643253184854985, + "grad_norm": 0.21733051538467407, + "learning_rate": 2.6274598520959966e-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": 0.010999999195337295, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.004985643154941499, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.026985641196370126, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.023643253184854985, + "weighted_rkd_loss": 0.013492820598185063 + }, + { + "epoch": 2.3076923076923075, + "flow_matching_loss": 0.022906559705734252, + "grad_norm": 0.28345248103141785, + "learning_rate": 2.6208917292491185e-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": 0.011069928156211972, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.005206426791846752, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.027346283197402954, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.022906559705734252, + "weighted_rkd_loss": 0.013673141598701477 + }, + { + "epoch": 2.324786324786325, + "flow_matching_loss": 0.02266085259616375, + "grad_norm": 0.1736982762813568, + "learning_rate": 2.6142745749668176e-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": 0.011071230052039027, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.004988636053167283, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.02713109627366066, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.02266085259616375, + "weighted_rkd_loss": 0.01356554813683033 + }, + { + "epoch": 2.341880341880342, + "flow_matching_loss": 0.02306621465831995, + "grad_norm": 0.26756277680397034, + "learning_rate": 2.607608678704859e-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": 0.010481357015669346, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.004531278763897717, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.0254939928650856, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.02306621465831995, + "weighted_rkd_loss": 0.0127469964325428 + }, + { + "epoch": 2.358974358974359, + "flow_matching_loss": 0.02586138006299734, + "grad_norm": 0.24713675677776337, + "learning_rate": 2.600894332051139e-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": 0.011580483987927437, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.004711030353792012, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.027871998213231563, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.02586138006299734, + "weighted_rkd_loss": 0.013935999106615781 + }, + { + "epoch": 2.376068376068376, + "flow_matching_loss": 0.026022879406809807, + "grad_norm": 0.21917226910591125, + "learning_rate": 2.594131828712932e-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": 0.011406512837857008, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.004805121943354607, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.027618147805333136, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.026022879406809807, + "weighted_rkd_loss": 0.013809073902666568 + }, + { + "epoch": 2.393162393162393, + "flow_matching_loss": 0.023596177250146864, + "grad_norm": 0.2035856992006302, + "learning_rate": 2.5873214645040414e-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": 0.010142395226284861, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0045007105683907865, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.024785500764846802, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.023596177250146864, + "weighted_rkd_loss": 0.012392750382423401 + }, + { + "epoch": 2.41025641025641, + "flow_matching_loss": 0.02610907070338726, + "grad_norm": 0.21590881049633026, + "learning_rate": 2.5804635373318604e-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": 0.008909029187634587, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0037137356353923678, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.02153179394081235, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.02610907070338726, + "weighted_rkd_loss": 0.010765896970406174 + }, + { + "epoch": 2.427350427350427, + "flow_matching_loss": 0.02329109571874142, + "grad_norm": 0.2922363579273224, + "learning_rate": 2.57355834718434e-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": 0.011788456933572888, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.004883810528554022, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.02846072446554899, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.02329109571874142, + "weighted_rkd_loss": 0.014230362232774496 + }, + { + "epoch": 2.4444444444444446, + "flow_matching_loss": 0.024115246534347535, + "grad_norm": 0.2587732970714569, + "learning_rate": 2.566606196116867e-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": 0.013214311562478542, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.005839744093827903, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.03226836714893579, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.024115246534347535, + "weighted_rkd_loss": 0.016134183574467897 + }, + { + "epoch": 2.4615384615384617, + "flow_matching_loss": 0.02489748504012823, + "grad_norm": 0.21666760742664337, + "learning_rate": 2.559607388239051e-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": 0.010760731622576713, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.004546113940887153, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.026067577116191388, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.02489748504012823, + "weighted_rkd_loss": 0.013033788558095694 + }, + { + "epoch": 2.4786324786324787, + "flow_matching_loss": 0.020900059305131436, + "grad_norm": 0.2407768964767456, + "learning_rate": 2.552562229701422e-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": 0.010831856867298484, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.004969510156661272, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.02663322389125824, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.020900059305131436, + "weighted_rkd_loss": 0.01331661194562912 + }, + { + "epoch": 2.4957264957264957, + "flow_matching_loss": 0.022048553638160228, + "grad_norm": 0.16484962403774261, + "learning_rate": 2.545471028682036e-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": 0.008939298242330551, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.003671724908053875, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.02155032129958272, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.022048553638160228, + "weighted_rkd_loss": 0.01077516064979136 + }, + { + "epoch": 2.5128205128205128, + "flow_matching_loss": 0.024487999454140662, + "grad_norm": 0.1850469410419464, + "learning_rate": 2.538334095373e-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": 0.010523739224299788, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.00469970719423145, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.025747185945510863, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.024487999454140662, + "weighted_rkd_loss": 0.012873592972755431 + }, + { + "epoch": 2.52991452991453, + "flow_matching_loss": 0.023977626115083694, + "grad_norm": 0.26976409554481506, + "learning_rate": 2.531151741966896e-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": 0.01056251721456647, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.005071718548424542, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.026196752954274417, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.023977626115083694, + "weighted_rkd_loss": 0.013098376477137208 + }, + { + "epoch": 2.547008547008547, + "flow_matching_loss": 0.02404904179275036, + "grad_norm": 0.22599978744983673, + "learning_rate": 2.5239242826431302e-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": 0.010430255578830839, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.004550645523704588, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.02541115628555417, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.02404904179275036, + "weighted_rkd_loss": 0.012705578142777086 + }, + { + "epoch": 2.564102564102564, + "flow_matching_loss": 0.026297237910330295, + "grad_norm": 0.2992944121360779, + "learning_rate": 2.5166520335541865e-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": 0.011783476080745458, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0052417038241401315, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.02880865577608347, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.026297237910330295, + "weighted_rkd_loss": 0.014404327888041735 + }, + { + "epoch": 2.5811965811965814, + "flow_matching_loss": 0.023364758491516112, + "grad_norm": 0.31836071610450745, + "learning_rate": 2.5093353128117996e-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": 0.010980317369103432, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.005202659475617111, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.02716329451650381, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.023364758491516112, + "weighted_rkd_loss": 0.013581647258251905 + }, + { + "epoch": 2.5982905982905984, + "flow_matching_loss": 0.022494507301598788, + "grad_norm": 0.2116675227880478, + "learning_rate": 2.501974440473037e-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": 0.010303075471892954, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.004565139859914779, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.025171290896832944, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.022494507301598788, + "weighted_rkd_loss": 0.012585645448416472 + }, + { + "epoch": 2.6153846153846154, + "flow_matching_loss": 0.026725856214761735, + "grad_norm": 0.24244584143161774, + "learning_rate": 2.4945697385263015e-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": 0.011164265777915716, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.005043569020926952, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.027372100763022898, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.026725856214761735, + "weighted_rkd_loss": 0.013686050381511449 + }, + { + "epoch": 2.6324786324786325, + "flow_matching_loss": 0.02198776612058282, + "grad_norm": 0.22380423545837402, + "learning_rate": 2.487121530877243e-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": 0.010480466578155757, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.004462654585950077, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.025423587858676912, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.02198776612058282, + "weighted_rkd_loss": 0.012711793929338456 + }, + { + "epoch": 2.6495726495726495, + "flow_matching_loss": 0.02080122269690037, + "grad_norm": 0.24043220281600952, + "learning_rate": 2.4796301433345943e-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": 0.008758699102327228, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0038610855117440224, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.02137848371639848, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.02080122269690037, + "weighted_rkd_loss": 0.01068924185819924 + }, + { + "epoch": 2.6666666666666665, + "flow_matching_loss": 0.023161404207348822, + "grad_norm": 0.24361535906791687, + "learning_rate": 2.4720959035959148e-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": 0.011636437475681305, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.005215252633206546, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.028488127514719963, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.023161404207348822, + "weighted_rkd_loss": 0.014244063757359982 + }, + { + "epoch": 2.683760683760684, + "flow_matching_loss": 0.02135396469384432, + "grad_norm": 0.24413393437862396, + "learning_rate": 2.464519141233257e-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": 0.010051483660936356, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0042491775238886476, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.024352144729346038, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.02135396469384432, + "weighted_rkd_loss": 0.012176072364673019 + }, + { + "epoch": 2.700854700854701, + "flow_matching_loss": 0.022891308926045895, + "grad_norm": 0.2563834488391876, + "learning_rate": 2.456900187678754e-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": 0.009299258328974247, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.004032260063104332, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.02263077683746815, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.022891308926045895, + "weighted_rkd_loss": 0.011315388418734074 + }, + { + "epoch": 2.717948717948718, + "flow_matching_loss": 0.019574376195669173, + "grad_norm": 0.2142282873392105, + "learning_rate": 2.4492393762101147e-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": 0.01088754371739924, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.005167186912149191, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.02694227434694767, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.019574376195669173, + "weighted_rkd_loss": 0.013471137173473835 + }, + { + "epoch": 2.735042735042735, + "flow_matching_loss": 0.023829315975308418, + "grad_norm": 0.2677934169769287, + "learning_rate": 2.4415370419360508e-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": 0.010912666656076908, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.004734570253640413, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.026559903100132943, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.023829315975308418, + "weighted_rkd_loss": 0.013279951550066471 + }, + { + "epoch": 2.752136752136752, + "flow_matching_loss": 0.022046511620283128, + "grad_norm": 0.18143510818481445, + "learning_rate": 2.4337935217816157e-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": 0.009173155901953578, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.004238010593689978, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.022584322560578584, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.022046511620283128, + "weighted_rkd_loss": 0.011292161280289292 + }, + { + "epoch": 2.769230769230769, + "flow_matching_loss": 0.0256638266146183, + "grad_norm": 0.26739150285720825, + "learning_rate": 2.4260091544734665e-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": 0.009626912791281939, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0045542302541434765, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.02380805565044284, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.0256638266146183, + "weighted_rkd_loss": 0.01190402782522142 + }, + { + "epoch": 2.786324786324786, + "flow_matching_loss": 0.02314116433262825, + "grad_norm": 0.27141332626342773, + "learning_rate": 2.4181842805250466e-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": 0.00963348699733615, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0044304505223408345, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.02369742449373007, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.02314116433262825, + "weighted_rkd_loss": 0.011848712246865034 + }, + { + "epoch": 2.8034188034188032, + "flow_matching_loss": 0.02181963212788105, + "grad_norm": 0.20872725546360016, + "learning_rate": 2.4103192422216923e-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": 0.009032536670565606, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.004314871900714934, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.022379945125430824, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.02181963212788105, + "weighted_rkd_loss": 0.011189972562715412 + }, + { + "epoch": 2.8205128205128203, + "flow_matching_loss": 0.022701138444244862, + "grad_norm": 0.1778896450996399, + "learning_rate": 2.402414383605658e-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": 0.010498900059610605, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0045100216753780845, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.02550782198086381, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.022701138444244862, + "weighted_rkd_loss": 0.012753910990431905 + }, + { + "epoch": 2.8376068376068377, + "flow_matching_loss": 0.024045399948954584, + "grad_norm": 0.25071728229522705, + "learning_rate": 2.3944700504610674e-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": 0.00897721559740603, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.003891855734400451, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.021846286952495575, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.024045399948954584, + "weighted_rkd_loss": 0.010923143476247787 + }, + { + "epoch": 2.8547008547008548, + "flow_matching_loss": 0.02234038049355149, + "grad_norm": 0.20470502972602844, + "learning_rate": 2.3864865902987883e-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": 0.011603549402207137, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0051022290484979745, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.028309327736496925, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.02234038049355149, + "weighted_rkd_loss": 0.014154663868248462 + }, + { + "epoch": 2.871794871794872, + "flow_matching_loss": 0.023738479986786843, + "grad_norm": 0.30578407645225525, + "learning_rate": 2.378464352341231e-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": 0.008975940523669124, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.003991176700219512, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.02194305770099163, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.023738479986786843, + "weighted_rkd_loss": 0.010971528850495816 + }, + { + "epoch": 2.888888888888889, + "flow_matching_loss": 0.022554483730345964, + "grad_norm": 0.1951245367527008, + "learning_rate": 2.3704036875070728e-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": 0.009812219673767686, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.004091479280032217, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.023715918976813556, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.022554483730345964, + "weighted_rkd_loss": 0.011857959488406778 + }, + { + "epoch": 2.905982905982906, + "flow_matching_loss": 0.02242042701691389, + "grad_norm": 0.2974187135696411, + "learning_rate": 2.362304948395906e-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": 0.010304574249312281, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.004178128391504288, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.024787276796996594, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.02242042701691389, + "weighted_rkd_loss": 0.012393638398498297 + }, + { + "epoch": 2.9230769230769234, + "flow_matching_loss": 0.023922652564942838, + "grad_norm": 0.18772614002227783, + "learning_rate": 2.3541684892728157e-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": 0.009445326868444681, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.004097510641440749, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.022988164238631724, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.023922652564942838, + "weighted_rkd_loss": 0.011494082119315862 + }, + { + "epoch": 2.9401709401709404, + "flow_matching_loss": 0.026682638563215733, + "grad_norm": 0.2330038994550705, + "learning_rate": 2.345994666052883e-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": 0.009895666409283877, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.00440065455622971, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.024191987700760365, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.026682638563215733, + "weighted_rkd_loss": 0.012095993850380182 + }, + { + "epoch": 2.9572649572649574, + "flow_matching_loss": 0.02124050781130791, + "grad_norm": 0.19767366349697113, + "learning_rate": 2.337783836285614e-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": 0.008275231532752513, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.003853382426314056, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.020403845328837635, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.02124050781130791, + "weighted_rkd_loss": 0.010201922664418817 + }, + { + "epoch": 2.9743589743589745, + "flow_matching_loss": 0.02081846250221133, + "grad_norm": 0.307229608297348, + "learning_rate": 2.3295363591393026e-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": 0.009578318148851395, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0040630410658195615, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.02321967724710703, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.02081846250221133, + "weighted_rkd_loss": 0.011609838623553515 + }, + { + "epoch": 2.9914529914529915, + "flow_matching_loss": 0.0220365384593606, + "grad_norm": 0.19888272881507874, + "learning_rate": 2.3212525953853174e-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": 0.009073050040751696, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.003744899039156735, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.021890999004244804, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.0220365384593606, + "weighted_rkd_loss": 0.010945499502122402 + }, + { + "epoch": 3.0085470085470085, + "flow_matching_loss": 0.020825703255832197, + "grad_norm": 0.1918894499540329, + "learning_rate": 2.3129329073823206e-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": 0.009859509905800224, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0045070439809933305, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.024226063676178456, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.020825703255832197, + "weighted_rkd_loss": 0.012113031838089228 + }, + { + "epoch": 3.0256410256410255, + "flow_matching_loss": 0.02404018212109804, + "grad_norm": 0.19095730781555176, + "learning_rate": 2.3045776590604175e-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": 0.008907624846324324, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.004010485555045307, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.021825735550373794, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.02404018212109804, + "weighted_rkd_loss": 0.010912867775186897 + }, + { + "epoch": 3.0427350427350426, + "flow_matching_loss": 0.02037848560139537, + "grad_norm": 0.23975929617881775, + "learning_rate": 2.296187215905237e-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": 0.009382487926632165, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0042326563969254495, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.022997632250189783, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.02037848560139537, + "weighted_rkd_loss": 0.011498816125094891 + }, + { + "epoch": 3.0598290598290596, + "flow_matching_loss": 0.02206662083044648, + "grad_norm": 0.14885295927524567, + "learning_rate": 2.2877619449419437e-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": 0.008644994953647256, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0035399044398218395, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.020829894207417966, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.02206662083044648, + "weighted_rkd_loss": 0.010414947103708983 + }, + { + "epoch": 3.076923076923077, + "flow_matching_loss": 0.020480931922793388, + "grad_norm": 0.20446527004241943, + "learning_rate": 2.2793022147191843e-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": 0.009374905563890933, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0042403008788824085, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.02299011219292879, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.020480931922793388, + "weighted_rkd_loss": 0.011495056096464396 + }, + { + "epoch": 3.094017094017094, + "flow_matching_loss": 0.021702952776104213, + "grad_norm": 0.19124248623847961, + "learning_rate": 2.2708083952929634e-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": 0.009283293923363089, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.004115194361656904, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.022681782208383083, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.021702952776104213, + "weighted_rkd_loss": 0.011340891104191541 + }, + { + "epoch": 3.111111111111111, + "flow_matching_loss": 0.023065047152340413, + "grad_norm": 0.16134604811668396, + "learning_rate": 2.2622808582104594e-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": 0.008247178699821233, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0038086207583546638, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.020302978157997132, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.023065047152340413, + "weighted_rkd_loss": 0.010151489078998566 + }, + { + "epoch": 3.128205128205128, + "flow_matching_loss": 0.020875094179064035, + "grad_norm": 0.21919263899326324, + "learning_rate": 2.2537199764937703e-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": 0.00895846332423389, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.00419829033780843, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.022115217056125402, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.020875094179064035, + "weighted_rkd_loss": 0.011057608528062701 + }, + { + "epoch": 3.1452991452991452, + "flow_matching_loss": 0.02138959923759103, + "grad_norm": 0.23678432404994965, + "learning_rate": 2.245126124623595e-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": 0.007867638859897852, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.003307641646824777, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.019042919389903545, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.02138959923759103, + "weighted_rkd_loss": 0.009521459694951773 + }, + { + "epoch": 3.1623931623931623, + "flow_matching_loss": 0.02178165577352047, + "grad_norm": 0.27445220947265625, + "learning_rate": 2.2364996785228547e-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": 0.008633937686681747, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.004018479073420167, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.021286354307085274, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.02178165577352047, + "weighted_rkd_loss": 0.010643177153542637 + }, + { + "epoch": 3.1794871794871793, + "flow_matching_loss": 0.021973855700343846, + "grad_norm": 0.2308157980442047, + "learning_rate": 2.227841015540248e-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": 0.007728042174130678, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0032928862725384533, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.018748970329761507, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.021973855700343846, + "weighted_rkd_loss": 0.009374485164880753 + }, + { + "epoch": 3.1965811965811968, + "flow_matching_loss": 0.021395375672727823, + "grad_norm": 0.21695679426193237, + "learning_rate": 2.2191505144337433e-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": 0.0077186715323477985, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0033147968002595007, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.018752139899879693, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.021395375672727823, + "weighted_rkd_loss": 0.009376069949939846 + }, + { + "epoch": 3.213675213675214, + "flow_matching_loss": 0.021534157823771237, + "grad_norm": 0.1653919219970703, + "learning_rate": 2.210428555354013e-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": 0.00908505842089653, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0035650128265842795, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.02173512978479266, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.021534157823771237, + "weighted_rkd_loss": 0.01086756489239633 + }, + { + "epoch": 3.230769230769231, + "flow_matching_loss": 0.02106862384825945, + "grad_norm": 0.2326592653989792, + "learning_rate": 2.201675519827802e-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": 0.009765352075919508, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.004243217711336911, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.023773921933025122, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.02106862384825945, + "weighted_rkd_loss": 0.011886960966512561 + }, + { + "epoch": 3.247863247863248, + "flow_matching_loss": 0.021206633374094962, + "grad_norm": 0.19293636083602905, + "learning_rate": 2.1928917907412403e-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": 0.007537059253081679, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.003404276235960424, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.018478394765406848, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.021206633374094962, + "weighted_rkd_loss": 0.009239197382703424 + }, + { + "epoch": 3.264957264957265, + "flow_matching_loss": 0.02262264396995306, + "grad_norm": 0.20528379082679749, + "learning_rate": 2.1840777523230937e-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": 0.008054467104375362, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0038436053320765494, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.019952539540827274, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.02262264396995306, + "weighted_rkd_loss": 0.009976269770413637 + }, + { + "epoch": 3.282051282051282, + "flow_matching_loss": 0.022744439914822577, + "grad_norm": 0.18302470445632935, + "learning_rate": 2.1752337901279552e-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": 0.011205578362569213, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.004998157313093543, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.027409313805401327, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.022744439914822577, + "weighted_rkd_loss": 0.013704656902700663 + }, + { + "epoch": 3.299145299145299, + "flow_matching_loss": 0.021091637015342713, + "grad_norm": 0.27177563309669495, + "learning_rate": 2.166360291019383e-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": 0.008100183354690671, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0032902485225349666, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.019490615092217924, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.021091637015342713, + "weighted_rkd_loss": 0.009745307546108962 + }, + { + "epoch": 3.316239316239316, + "flow_matching_loss": 0.022644120268523693, + "grad_norm": 0.20682282745838165, + "learning_rate": 2.157457643152973e-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": 0.009497429803013801, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.00464288117364049, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.02363774077966809, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.022644120268523693, + "weighted_rkd_loss": 0.011818870389834046 + }, + { + "epoch": 3.3333333333333335, + "flow_matching_loss": 0.02207537516951561, + "grad_norm": 0.24180874228477478, + "learning_rate": 2.148526235959385e-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": 0.008376868302002549, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0032720234477892516, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.020025760307908057, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.02207537516951561, + "weighted_rkd_loss": 0.010012880153954028 + }, + { + "epoch": 3.3504273504273505, + "flow_matching_loss": 0.02177107408642769, + "grad_norm": 0.25179964303970337, + "learning_rate": 2.1395664601273016e-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": 0.00934967128559947, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.004402558598667383, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.023101901076734065, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.02177107408642769, + "weighted_rkd_loss": 0.011550950538367032 + }, + { + "epoch": 3.3675213675213675, + "flow_matching_loss": 0.021467326395213603, + "grad_norm": 0.20472826063632965, + "learning_rate": 2.130578707586345e-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": 0.00802918765693903, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.00377117128809914, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01982954628765583, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.021467326395213603, + "weighted_rkd_loss": 0.009914773143827915 + }, + { + "epoch": 3.3846153846153846, + "flow_matching_loss": 0.022415131144225598, + "grad_norm": 0.18386898934841156, + "learning_rate": 2.1215633714899264e-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": 0.008150833193212748, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0038708670064806936, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.020172533206641675, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.022415131144225598, + "weighted_rkd_loss": 0.010086266603320838 + }, + { + "epoch": 3.4017094017094016, + "flow_matching_loss": 0.02110731778666377, + "grad_norm": 0.1889202892780304, + "learning_rate": 2.112520846198053e-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": 0.00840694373473525, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.003788764402270317, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.020602651871740818, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.02110731778666377, + "weighted_rkd_loss": 0.010301325935870409 + }, + { + "epoch": 3.4188034188034186, + "flow_matching_loss": 0.019866282120347023, + "grad_norm": 0.26760584115982056, + "learning_rate": 2.103451527260074e-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": 0.011116993194445968, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0044356569647789005, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.02666964316740632, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.019866282120347023, + "weighted_rkd_loss": 0.01333482158370316 + }, + { + "epoch": 3.435897435897436, + "flow_matching_loss": 0.018249011039733885, + "grad_norm": 0.23657004535198212, + "learning_rate": 2.094355811397381e-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": 0.008862052159383894, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0038880287203937767, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.021612132992595436, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.018249011039733885, + "weighted_rkd_loss": 0.010806066496297718 + }, + { + "epoch": 3.452991452991453, + "flow_matching_loss": 0.022592077776789666, + "grad_norm": 0.16656915843486786, + "learning_rate": 2.0852340964860522e-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": 0.008196860598400236, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0035904660122469067, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.019984187465161084, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.022592077776789666, + "weighted_rkd_loss": 0.009992093732580542 + }, + { + "epoch": 3.47008547008547, + "flow_matching_loss": 0.021013226732611655, + "grad_norm": 0.2137608677148819, + "learning_rate": 2.076086781539447e-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": 0.008774524834007025, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.004004479385912418, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.021553528774529696, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.021013226732611655, + "weighted_rkd_loss": 0.010776764387264848 + }, + { + "epoch": 3.4871794871794872, + "flow_matching_loss": 0.019182110857218503, + "grad_norm": 0.2665335237979889, + "learning_rate": 2.066914266690755e-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": 0.008432052610442043, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0039183218264952305, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.020782427024096252, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.019182110857218503, + "weighted_rkd_loss": 0.010391213512048126 + }, + { + "epoch": 3.5042735042735043, + "flow_matching_loss": 0.02038408424705267, + "grad_norm": 0.15456117689609528, + "learning_rate": 2.0577169531754906e-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": 0.010282731615006924, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.004464774415828288, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.025030237529426812, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.02038408424705267, + "weighted_rkd_loss": 0.012515118764713406 + }, + { + "epoch": 3.5213675213675213, + "flow_matching_loss": 0.021545465849339963, + "grad_norm": 0.1929696649312973, + "learning_rate": 2.048495243313942e-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": 0.008824501093477011, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0036449615377932786, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.02129396367818117, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.021545465849339963, + "weighted_rkd_loss": 0.010646981839090585 + }, + { + "epoch": 3.5384615384615383, + "flow_matching_loss": 0.018455321434885263, + "grad_norm": 0.23541535437107086, + "learning_rate": 2.039249540493572e-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": 0.00999117074534297, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.003966285148635507, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.023948626406490803, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.018455321434885263, + "weighted_rkd_loss": 0.011974313203245402 + }, + { + "epoch": 3.5555555555555554, + "flow_matching_loss": 0.018786207027733325, + "grad_norm": 0.22462044656276703, + "learning_rate": 2.0299802491513755e-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": 0.00863970723003149, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0035249478416517375, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.02080436246469617, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.018786207027733325, + "weighted_rkd_loss": 0.010402181232348085 + }, + { + "epoch": 3.5726495726495724, + "flow_matching_loss": 0.019480399787425995, + "grad_norm": 0.2128298580646515, + "learning_rate": 2.0206877747561828e-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": 0.007677402580156922, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.003288325946778059, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01864313129335642, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.019480399787425995, + "weighted_rkd_loss": 0.00932156564667821 + }, + { + "epoch": 3.58974358974359, + "flow_matching_loss": 0.01911817295476794, + "grad_norm": 0.23664461076259613, + "learning_rate": 2.011372523790927e-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": 0.00922760833054781, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.004206936457194388, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.022662153095006944, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.01911817295476794, + "weighted_rkd_loss": 0.011331076547503472 + }, + { + "epoch": 3.606837606837607, + "flow_matching_loss": 0.021250076405704022, + "grad_norm": 0.21973693370819092, + "learning_rate": 2.0020349037348643e-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": 0.007971875090152025, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.003988684457726777, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.019932434521615507, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.021250076405704022, + "weighted_rkd_loss": 0.009966217260807753 + }, + { + "epoch": 3.623931623931624, + "flow_matching_loss": 0.019656579475849867, + "grad_norm": 0.21634699404239655, + "learning_rate": 1.9926753230457443e-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": 0.00955271115526557, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.003916142298839987, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.023021564818918705, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.019656579475849867, + "weighted_rkd_loss": 0.011510782409459352 + }, + { + "epoch": 3.641025641025641, + "flow_matching_loss": 0.02276723589748144, + "grad_norm": 0.15986588597297668, + "learning_rate": 1.983294191141948e-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": 0.0075195180252194405, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.003291030717082322, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01833006674423814, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.02276723589748144, + "weighted_rkd_loss": 0.00916503337211907 + }, + { + "epoch": 3.658119658119658, + "flow_matching_loss": 0.020830532629042865, + "grad_norm": 0.21286843717098236, + "learning_rate": 1.973891918384575e-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": 0.008494880562648178, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.003529363591223955, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.020519124809652567, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.020830532629042865, + "weighted_rkd_loss": 0.010259562404826283 + }, + { + "epoch": 3.6752136752136755, + "flow_matching_loss": 0.023566334135830404, + "grad_norm": 0.24354811012744904, + "learning_rate": 1.9644689160594945e-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": 0.008519923128187657, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0036325100576505066, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.02067235643044114, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.023566334135830404, + "weighted_rkd_loss": 0.01033617821522057 + }, + { + "epoch": 3.6923076923076925, + "flow_matching_loss": 0.018900660425424577, + "grad_norm": 0.1920747309923172, + "learning_rate": 1.9550255963593553e-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": 0.007615871122106909, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.003055795456748456, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.018287537898868323, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.018900660425424577, + "weighted_rkd_loss": 0.009143768949434162 + }, + { + "epoch": 3.7094017094017095, + "flow_matching_loss": 0.01953559760004282, + "grad_norm": 0.19230477511882782, + "learning_rate": 1.9455623723655524e-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": 0.0067968220915645364, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0031586338998749854, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.016752277966588736, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.01953559760004282, + "weighted_rkd_loss": 0.008376138983294368 + }, + { + "epoch": 3.7264957264957266, + "flow_matching_loss": 0.020689549297094344, + "grad_norm": 0.23367641866207123, + "learning_rate": 1.9360796580301606e-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": 0.008445434179157019, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.003690579044632614, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.020581447705626487, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.020689549297094344, + "weighted_rkd_loss": 0.010290723852813244 + }, + { + "epoch": 3.7435897435897436, + "flow_matching_loss": 0.020847134850919248, + "grad_norm": 0.2316247820854187, + "learning_rate": 1.9265778681578245e-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": 0.008731900528073312, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0034508655546233056, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.020914666634052993, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.020847134850919248, + "weighted_rkd_loss": 0.010457333317026496 + }, + { + "epoch": 3.7606837606837606, + "flow_matching_loss": 0.02032330334186554, + "grad_norm": 0.17385540902614594, + "learning_rate": 1.9170574183876143e-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": 0.007317671971395612, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0032924892380833625, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.017927832994610073, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.02032330334186554, + "weighted_rkd_loss": 0.008963916497305036 + }, + { + "epoch": 3.7777777777777777, + "flow_matching_loss": 0.016899863351136447, + "grad_norm": 0.26192307472229004, + "learning_rate": 1.9075187251748452e-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": 0.007569550909101963, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.003170735482126474, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.018309837207198143, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.016899863351136447, + "weighted_rkd_loss": 0.009154918603599072 + }, + { + "epoch": 3.7948717948717947, + "flow_matching_loss": 0.018195388186722995, + "grad_norm": 0.25618091225624084, + "learning_rate": 1.8979622057728596e-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": 0.007942789187654852, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0032951420871540904, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.019180720765143634, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.018195388186722995, + "weighted_rkd_loss": 0.009590360382571817 + }, + { + "epoch": 3.8119658119658117, + "flow_matching_loss": 0.02252973020076752, + "grad_norm": 0.1578928828239441, + "learning_rate": 1.8883882782147748e-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": 0.0074423910584300755, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.003065627347677946, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.017950409650802614, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.02252973020076752, + "weighted_rkd_loss": 0.008975204825401307 + }, + { + "epoch": 3.8290598290598292, + "flow_matching_loss": 0.02154372911900282, + "grad_norm": 0.20821666717529297, + "learning_rate": 1.8787973612951967e-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": 0.00786780584603548, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0032656229101121425, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01900123441591859, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.02154372911900282, + "weighted_rkd_loss": 0.009500617207959295 + }, + { + "epoch": 3.8461538461538463, + "flow_matching_loss": 0.020391280949115752, + "grad_norm": 0.17726753652095795, + "learning_rate": 1.8691898745519028e-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": 0.007469217339530587, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.003028665063902736, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01796709969639778, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.020391280949115752, + "weighted_rkd_loss": 0.00898354984819889 + }, + { + "epoch": 3.8632478632478633, + "flow_matching_loss": 0.02223206451162696, + "grad_norm": 0.21495142579078674, + "learning_rate": 1.8595662382474863e-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": 0.007301053707487881, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.003117244434542954, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.017719351686537266, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.02223206451162696, + "weighted_rkd_loss": 0.008859675843268633 + }, + { + "epoch": 3.8803418803418803, + "flow_matching_loss": 0.01902555041015148, + "grad_norm": 0.20775483548641205, + "learning_rate": 1.8499268733509766e-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": 0.0073735327925533055, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.003244478744454682, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.017991544399410486, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.01902555041015148, + "weighted_rkd_loss": 0.008995772199705243 + }, + { + "epoch": 3.8974358974358974, + "flow_matching_loss": 0.021454056072980166, + "grad_norm": 0.19884011149406433, + "learning_rate": 1.84027220151942e-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": 0.007698776153847575, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0034163655480369927, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.018813917879015207, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.021454056072980166, + "weighted_rkd_loss": 0.009406958939507604 + }, + { + "epoch": 3.9145299145299144, + "flow_matching_loss": 0.020456366799771787, + "grad_norm": 0.21196779608726501, + "learning_rate": 1.83060264507944e-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": 0.007814684696495534, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0031296714674681424, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.018759041093289852, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.020456366799771787, + "weighted_rkd_loss": 0.009379520546644926 + }, + { + "epoch": 3.931623931623932, + "flow_matching_loss": 0.019748034421354532, + "grad_norm": 0.21080656349658966, + "learning_rate": 1.82091862700876e-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": 0.007885579485446215, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0030657083028927446, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.018836867436766623, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.019748034421354532, + "weighted_rkd_loss": 0.009418433718383312 + }, + { + "epoch": 3.948717948717949, + "flow_matching_loss": 0.021026203967630865, + "grad_norm": 0.19430294632911682, + "learning_rate": 1.8112205709177023e-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": 0.007923238072544336, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0032659520395100115, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.019112428463995457, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.021026203967630865, + "weighted_rkd_loss": 0.009556214231997729 + }, + { + "epoch": 3.965811965811966, + "flow_matching_loss": 0.019944813381880522, + "grad_norm": 0.32070183753967285, + "learning_rate": 1.8015089010306585e-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": 0.009109126357361674, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.003980800183489919, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.022199052944779397, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.019944813381880522, + "weighted_rkd_loss": 0.011099526472389699 + }, + { + "epoch": 3.982905982905983, + "flow_matching_loss": 0.019468062836676837, + "grad_norm": 0.25614234805107117, + "learning_rate": 1.7917840421675298e-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": 0.008891282556578516, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.003779213014058769, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.021561777964234353, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.019468062836676837, + "weighted_rkd_loss": 0.010780888982117176 + }, + { + "epoch": 4.0, + "flow_matching_loss": 0.017012250050902365, + "grad_norm": 0.16915491223335266, + "learning_rate": 1.7820464197251474e-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": 0.006800329172983765, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0029427450615912678, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01654340336099267, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.017012250050902365, + "weighted_rkd_loss": 0.008271701680496336 + }, + { + "epoch": 4.017094017094017, + "flow_matching_loss": 0.023955489788204432, + "grad_norm": 0.22803570330142975, + "learning_rate": 1.7722964596586627e-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": 0.006895763799548149, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.002875331451650709, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.016666858922690153, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.023955489788204432, + "weighted_rkd_loss": 0.008333429461345077 + }, + { + "epoch": 4.034188034188034, + "flow_matching_loss": 0.017703853361308575, + "grad_norm": 0.19360429048538208, + "learning_rate": 1.7625345884629144e-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": 0.007683675503358245, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.003063351823948324, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.018430702853947878, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.017703853361308575, + "weighted_rkd_loss": 0.009215351426973939 + }, + { + "epoch": 4.051282051282051, + "flow_matching_loss": 0.02146550826728344, + "grad_norm": 0.15965017676353455, + "learning_rate": 1.7527612331537722e-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": 0.0066555246710777284, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.002755923569202423, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.016066972725093366, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.02146550826728344, + "weighted_rkd_loss": 0.008033486362546683 + }, + { + "epoch": 4.068376068376068, + "flow_matching_loss": 0.01993013881146908, + "grad_norm": 0.20406599342823029, + "learning_rate": 1.7429768212494593e-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": 0.006303849536925555, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0027177043608389794, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.015325403492897748, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.01993013881146908, + "weighted_rkd_loss": 0.007662701746448874 + }, + { + "epoch": 4.085470085470085, + "flow_matching_loss": 0.01773031707853079, + "grad_norm": 0.19607993960380554, + "learning_rate": 1.7331817807518497e-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": 0.007189767109230161, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0032816751045174897, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.017661209404468536, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.01773031707853079, + "weighted_rkd_loss": 0.008830604702234268 + }, + { + "epoch": 4.102564102564102, + "flow_matching_loss": 0.018448451347649098, + "grad_norm": 0.1936192661523819, + "learning_rate": 1.723376540127745e-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": 0.006861878465861082, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0031465153908357022, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.016870272066444158, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.018448451347649098, + "weighted_rkd_loss": 0.008435136033222079 + }, + { + "epoch": 4.119658119658119, + "flow_matching_loss": 0.01903025442734361, + "grad_norm": 0.28160524368286133, + "learning_rate": 1.7135615282901357e-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": 0.007035943772643804, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.002800643094815314, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.016872530709952115, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.01903025442734361, + "weighted_rkd_loss": 0.008436265354976057 + }, + { + "epoch": 4.136752136752137, + "flow_matching_loss": 0.02104166429489851, + "grad_norm": 0.15420423448085785, + "learning_rate": 1.7037371745794358e-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": 0.008220721362158656, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0032787902280688288, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.019720232859253884, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.02104166429489851, + "weighted_rkd_loss": 0.009860116429626942 + }, + { + "epoch": 4.153846153846154, + "flow_matching_loss": 0.019651689007878302, + "grad_norm": 0.15077415108680725, + "learning_rate": 1.6939039087447026e-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": 0.0075453647645190355, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0031900717760436235, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.018280801270157097, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.019651689007878302, + "weighted_rkd_loss": 0.009140400635078549 + }, + { + "epoch": 4.170940170940171, + "flow_matching_loss": 0.020570564549416304, + "grad_norm": 0.15573781728744507, + "learning_rate": 1.6840621609248397e-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": 0.007544708624482155, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.003139568492770195, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.018228985741734503, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.020570564549416304, + "weighted_rkd_loss": 0.009114492870867252 + }, + { + "epoch": 4.188034188034188, + "flow_matching_loss": 0.019456946291029453, + "grad_norm": 0.20012281835079193, + "learning_rate": 1.6742123616297794e-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": 0.006495161354541779, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0027588428696617485, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01574916560202837, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.019456946291029453, + "weighted_rkd_loss": 0.007874582801014185 + }, + { + "epoch": 4.205128205128205, + "flow_matching_loss": 0.019614775478839875, + "grad_norm": 0.1647314727306366, + "learning_rate": 1.664354941721651e-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": 0.006208974169567227, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.002630527503788471, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01504847602918744, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.019614775478839875, + "weighted_rkd_loss": 0.00752423801459372 + }, + { + "epoch": 4.222222222222222, + "flow_matching_loss": 0.0180529972538352, + "grad_norm": 0.1961066573858261, + "learning_rate": 1.6544903323959355e-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": 0.0073069647885859014, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.002847653639037162, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.017461583390831946, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.0180529972538352, + "weighted_rkd_loss": 0.008730791695415973 + }, + { + "epoch": 4.239316239316239, + "flow_matching_loss": 0.020445212721824646, + "grad_norm": 0.22484345734119415, + "learning_rate": 1.6446189651626023e-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": 0.006985000707209111, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0031812989618629217, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.017151300329715014, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.020445212721824646, + "weighted_rkd_loss": 0.008575650164857507 + }, + { + "epoch": 4.256410256410256, + "flow_matching_loss": 0.019169483799487352, + "grad_norm": 0.2385309785604477, + "learning_rate": 1.6347412718272316e-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": 0.006840714206919074, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0025678937672637403, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.016249322239309548, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.019169483799487352, + "weighted_rkd_loss": 0.008124661119654774 + }, + { + "epoch": 4.273504273504273, + "flow_matching_loss": 0.017080207541584968, + "grad_norm": 0.17671377956867218, + "learning_rate": 1.624857684472131e-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": 0.007922574994154274, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.002863072499167174, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.018708222638815643, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.017080207541584968, + "weighted_rkd_loss": 0.009354111319407821 + }, + { + "epoch": 4.2905982905982905, + "flow_matching_loss": 0.019747746549546717, + "grad_norm": 0.24573016166687012, + "learning_rate": 1.6149686354374283e-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": 0.007329531060531735, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0030151148559525607, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.017674177046865226, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.019747746549546717, + "weighted_rkd_loss": 0.008837088523432613 + }, + { + "epoch": 4.3076923076923075, + "flow_matching_loss": 0.018726011179387568, + "grad_norm": 0.22373729944229126, + "learning_rate": 1.605074557302165e-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": 0.006631332472898066, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.002862296684179455, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.016124961618334054, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.018726011179387568, + "weighted_rkd_loss": 0.008062480809167027 + }, + { + "epoch": 4.3247863247863245, + "flow_matching_loss": 0.021509300265461207, + "grad_norm": 0.1717628836631775, + "learning_rate": 1.59517588286537e-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": 0.006188838463276625, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0025393213145434856, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01491699842736125, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.021509300265461207, + "weighted_rkd_loss": 0.007458499213680625 + }, + { + "epoch": 4.3418803418803416, + "flow_matching_loss": 0.020329602248966693, + "grad_norm": 0.18526530265808105, + "learning_rate": 1.5852730451271324e-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": 0.007229596329852939, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.002875377354212105, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01733456999063492, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.020329602248966693, + "weighted_rkd_loss": 0.00866728499531746 + }, + { + "epoch": 4.358974358974359, + "flow_matching_loss": 0.0184131451882422, + "grad_norm": 0.18414761126041412, + "learning_rate": 1.5753664772696546e-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": 0.006622652034275233, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0026653855573385956, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01591068971902132, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.0184131451882422, + "weighted_rkd_loss": 0.00795534485951066 + }, + { + "epoch": 4.3760683760683765, + "flow_matching_loss": 0.02079639583826065, + "grad_norm": 0.18560418486595154, + "learning_rate": 1.565456612638307e-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": 0.007964245975017548, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.003640728979371488, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.019569220952689648, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.02079639583826065, + "weighted_rkd_loss": 0.009784610476344824 + }, + { + "epoch": 4.3931623931623935, + "flow_matching_loss": 0.01935685221105814, + "grad_norm": 0.20377367734909058, + "learning_rate": 1.5555438847226732e-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": 0.0062549715396016834, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0029534086119383574, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.015463351644575596, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.01935685221105814, + "weighted_rkd_loss": 0.007731675822287798 + }, + { + "epoch": 4.410256410256411, + "flow_matching_loss": 0.018714569322764874, + "grad_norm": 0.16832947731018066, + "learning_rate": 1.545628727137584e-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": 0.006530326465144754, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0027387462090700866, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.015799399092793466, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.018714569322764874, + "weighted_rkd_loss": 0.007899699546396733 + }, + { + "epoch": 4.427350427350428, + "flow_matching_loss": 0.019943490531295537, + "grad_norm": 0.17356619238853455, + "learning_rate": 1.535711573604153e-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": 0.006290619866922498, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0028495809994637964, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.015430820826441049, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.019943490531295537, + "weighted_rkd_loss": 0.0077154104132205244 + }, + { + "epoch": 4.444444444444445, + "flow_matching_loss": 0.017166621424257754, + "grad_norm": 0.1952560544013977, + "learning_rate": 1.5257928579308032e-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": 0.006391754373908043, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0028482529683969914, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.015631761588156223, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.017166621424257754, + "weighted_rkd_loss": 0.007815880794078111 + }, + { + "epoch": 4.461538461538462, + "flow_matching_loss": 0.017560683749616147, + "grad_norm": 0.1988353580236435, + "learning_rate": 1.5158730139942902e-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": 0.007574155880138278, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0033369460492394865, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.018485257867723703, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.017560683749616147, + "weighted_rkd_loss": 0.009242628933861852 + }, + { + "epoch": 4.478632478632479, + "flow_matching_loss": 0.019688459113240242, + "grad_norm": 0.19683468341827393, + "learning_rate": 1.5059524757207239e-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": 0.0063693718519061806, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0031602281494997444, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.015898971911519767, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.019688459113240242, + "weighted_rkd_loss": 0.007949485955759884 + }, + { + "epoch": 4.495726495726496, + "flow_matching_loss": 0.01867778431624174, + "grad_norm": 0.2284543812274933, + "learning_rate": 1.4960316770665866e-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": 0.006279498664662242, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0026411876315250993, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.015200185030698777, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.01867778431624174, + "weighted_rkd_loss": 0.007600092515349388 + }, + { + "epoch": 4.512820512820513, + "flow_matching_loss": 0.01726768594235182, + "grad_norm": 0.1803433746099472, + "learning_rate": 1.4861110519997495e-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": 0.005782140232622623, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0024866451509296893, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.014050925616174936, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.01726768594235182, + "weighted_rkd_loss": 0.007025462808087468 + }, + { + "epoch": 4.52991452991453, + "flow_matching_loss": 0.018991167843341827, + "grad_norm": 0.24880416691303253, + "learning_rate": 1.4761910344804913e-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": 0.008187596360221506, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.00336456069489941, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01973975347355008, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.018991167843341827, + "weighted_rkd_loss": 0.00986987673677504 + }, + { + "epoch": 4.547008547008547, + "flow_matching_loss": 0.020480296481400727, + "grad_norm": 0.15083114802837372, + "learning_rate": 1.466272058442516e-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": 0.006214395631104708, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0026107010897248985, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.015039492212235928, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.020480296481400727, + "weighted_rkd_loss": 0.007519746106117964 + }, + { + "epoch": 4.564102564102564, + "flow_matching_loss": 0.016016822867095472, + "grad_norm": 0.24289827048778534, + "learning_rate": 1.4563545577739676e-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": 0.007098955847322941, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.003096250956878066, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.017294162698090077, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.016016822867095472, + "weighted_rkd_loss": 0.008647081349045039 + }, + { + "epoch": 4.581196581196581, + "flow_matching_loss": 0.016501933615654708, + "grad_norm": 0.23007218539714813, + "learning_rate": 1.4464389662984543e-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": 0.0073565777856856585, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0026881229132413864, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.017401278391480444, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.016501933615654708, + "weighted_rkd_loss": 0.008700639195740222 + }, + { + "epoch": 4.598290598290598, + "flow_matching_loss": 0.016896218247711658, + "grad_norm": 0.2506307065486908, + "learning_rate": 1.4365257177560684e-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": 0.00605305025819689, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.002633159328252077, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.014739259891211987, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.016896218247711658, + "weighted_rkd_loss": 0.007369629945605994 + }, + { + "epoch": 4.615384615384615, + "flow_matching_loss": 0.019083459116518497, + "grad_norm": 0.1666194647550583, + "learning_rate": 1.426615245784416e-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": 0.006203994457609952, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0027952198404818775, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.015203209035098553, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.019083459116518497, + "weighted_rkd_loss": 0.0076016045175492765 + }, + { + "epoch": 4.632478632478632, + "flow_matching_loss": 0.017737826984375715, + "grad_norm": 0.20131826400756836, + "learning_rate": 1.4167079838996468e-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": 0.005915038753300905, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0021665282547473907, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.013996605761349201, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.017737826984375715, + "weighted_rkd_loss": 0.006998302880674601 + }, + { + "epoch": 4.64957264957265, + "flow_matching_loss": 0.01762369042262435, + "grad_norm": 0.2097741961479187, + "learning_rate": 1.4068043654774897e-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": 0.0066073834663257, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0024939873139373956, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.0157087542116642, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.01762369042262435, + "weighted_rkd_loss": 0.0078543771058321 + }, + { + "epoch": 4.666666666666667, + "flow_matching_loss": 0.017110300622880457, + "grad_norm": 0.1796332746744156, + "learning_rate": 1.3969048237342992e-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": 0.006355493422597646, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.002593577420338988, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.015304564125835895, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.017110300622880457, + "weighted_rkd_loss": 0.007652282062917948 + }, + { + "epoch": 4.683760683760684, + "flow_matching_loss": 0.020454686228185893, + "grad_norm": 0.18637192249298096, + "learning_rate": 1.3870097917081011e-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": 0.0053040543105453255, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.002318077115342021, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.012926185876131058, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.020454686228185893, + "weighted_rkd_loss": 0.006463092938065529 + }, + { + "epoch": 4.700854700854701, + "flow_matching_loss": 0.016615460719913246, + "grad_norm": 0.18428291380405426, + "learning_rate": 1.3771197022396508e-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": 0.006608629692345857, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0031284614000469447, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.016345720645040272, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.016615460719913246, + "weighted_rkd_loss": 0.008172860322520136 + }, + { + "epoch": 4.717948717948718, + "flow_matching_loss": 0.018700057454407216, + "grad_norm": 0.15285761654376984, + "learning_rate": 1.3672349879535015e-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": 0.008110217843204736, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.003631356067489833, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.019851791579276322, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.018700057454407216, + "weighted_rkd_loss": 0.009925895789638161 + }, + { + "epoch": 4.735042735042735, + "flow_matching_loss": 0.016846880689263342, + "grad_norm": 0.18228141963481903, + "learning_rate": 1.3573560812390781e-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": 0.005839766724966467, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.002764455007854849, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01444398844614625, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.016846880689263342, + "weighted_rkd_loss": 0.007221994223073125 + }, + { + "epoch": 4.752136752136752, + "flow_matching_loss": 0.017821814958006145, + "grad_norm": 0.17568399012088776, + "learning_rate": 1.3474834142317635e-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": 0.006056840531527996, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0024255528813228013, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.014539233967661857, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.017821814958006145, + "weighted_rkd_loss": 0.007269616983830929 + }, + { + "epoch": 4.769230769230769, + "flow_matching_loss": 0.01895545031875372, + "grad_norm": 0.12359953671693802, + "learning_rate": 1.3376174187939955e-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": 0.006296288664452731, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0026273989700712264, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.015219976473599673, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.01895545031875372, + "weighted_rkd_loss": 0.0076099882367998365 + }, + { + "epoch": 4.786324786324786, + "flow_matching_loss": 0.01891577858477831, + "grad_norm": 0.15576958656311035, + "learning_rate": 1.3277585264963761e-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": 0.005916955787688494, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.002357321046292782, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.014191232435405255, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.01891577858477831, + "weighted_rkd_loss": 0.0070956162177026275 + }, + { + "epoch": 4.803418803418803, + "flow_matching_loss": 0.018333879485726356, + "grad_norm": 0.22275856137275696, + "learning_rate": 1.3179071685987928e-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": 0.006870065536350012, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0026655447203665973, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.016405675560235977, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.018333879485726356, + "weighted_rkd_loss": 0.008202837780117989 + }, + { + "epoch": 4.82051282051282, + "flow_matching_loss": 0.019350035302340984, + "grad_norm": 0.18743370473384857, + "learning_rate": 1.308063776031554e-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": 0.00618456348311156, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0021372218267060815, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.014506348874419928, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.019350035302340984, + "weighted_rkd_loss": 0.007253174437209964 + }, + { + "epoch": 4.837606837606837, + "flow_matching_loss": 0.016245046723634006, + "grad_norm": 0.20033948123455048, + "learning_rate": 1.2982287793765389e-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": 0.0066329317167401316, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0023788953665643932, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.015644758567214014, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.016245046723634006, + "weighted_rkd_loss": 0.007822379283607007 + }, + { + "epoch": 4.854700854700854, + "flow_matching_loss": 0.016420507896691562, + "grad_norm": 0.1553906500339508, + "learning_rate": 1.2884026088483616e-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": 0.0058491195552051066, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.002730893774423748, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.014429132873192429, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.016420507896691562, + "weighted_rkd_loss": 0.007214566436596214 + }, + { + "epoch": 4.871794871794872, + "flow_matching_loss": 0.021777248568832876, + "grad_norm": 0.17237120866775513, + "learning_rate": 1.2785856942755543e-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": 0.005849478789605201, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0024362468044273554, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.014135204302147032, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.021777248568832876, + "weighted_rkd_loss": 0.007067602151073516 + }, + { + "epoch": 4.888888888888889, + "flow_matching_loss": 0.018680552765727044, + "grad_norm": 0.19036276638507843, + "learning_rate": 1.2687784650817616e-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": 0.00545023640152067, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.002043917903210968, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.012944390717893838, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.018680552765727044, + "weighted_rkd_loss": 0.006472195358946919 + }, + { + "epoch": 4.905982905982906, + "flow_matching_loss": 0.0194564713165164, + "grad_norm": 0.18401499092578888, + "learning_rate": 1.2589813502669603e-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": 0.007179317437112332, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0028180433670058846, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01717667831107974, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.0194564713165164, + "weighted_rkd_loss": 0.00858833915553987 + }, + { + "epoch": 4.923076923076923, + "flow_matching_loss": 0.018249529600143432, + "grad_norm": 0.15625852346420288, + "learning_rate": 1.2491947783886907e-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": 0.005985748581588268, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0025901356828399004, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.014561632834374904, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.018249529600143432, + "weighted_rkd_loss": 0.007280816417187452 + }, + { + "epoch": 4.94017094017094, + "flow_matching_loss": 0.019586964324116708, + "grad_norm": 0.19170287251472473, + "learning_rate": 1.2394191775433106e-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": 0.006269854912534356, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0026000387500971557, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.015139748714864254, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.019586964324116708, + "weighted_rkd_loss": 0.007569874357432127 + }, + { + "epoch": 4.957264957264957, + "flow_matching_loss": 0.018550141993910075, + "grad_norm": 0.217694953083992, + "learning_rate": 1.2296549753472696e-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": 0.0063949782401323315, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0023791401297785343, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.015169096365571021, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.018550141993910075, + "weighted_rkd_loss": 0.007584548182785511 + }, + { + "epoch": 4.9743589743589745, + "flow_matching_loss": 0.01646477123722434, + "grad_norm": 0.1519484668970108, + "learning_rate": 1.2199025989184032e-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": 0.006408982444554567, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.00265109499450773, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.015469059906899929, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.01646477123722434, + "weighted_rkd_loss": 0.007734529953449964 + }, + { + "epoch": 4.9914529914529915, + "flow_matching_loss": 0.01712658563628793, + "grad_norm": 0.17936837673187256, + "learning_rate": 1.21016247485725e-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": 0.005534190079197287, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.002666365006007254, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01373474532738328, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.01712658563628793, + "weighted_rkd_loss": 0.00686737266369164 + }, + { + "epoch": 5.0085470085470085, + "flow_matching_loss": 0.017433699034154416, + "grad_norm": 0.19425033032894135, + "learning_rate": 1.2004350292283897e-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": 0.005776865407824516, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.002266028500162065, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.013819759339094162, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.017433699034154416, + "weighted_rkd_loss": 0.006909879669547081 + }, + { + "epoch": 5.0256410256410255, + "flow_matching_loss": 0.01623713644221425, + "grad_norm": 0.15402722358703613, + "learning_rate": 1.1907206875418068e-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": 0.005933634308166802, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0025461736135184766, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01441344222985208, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.01623713644221425, + "weighted_rkd_loss": 0.00720672111492604 + }, + { + "epoch": 5.042735042735043, + "flow_matching_loss": 0.01722178068011999, + "grad_norm": 0.17161601781845093, + "learning_rate": 1.1810198747342769e-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": 0.006133737275376916, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0025711448513902726, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.014838619343936443, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.01722178068011999, + "weighted_rkd_loss": 0.007419309671968221 + }, + { + "epoch": 5.05982905982906, + "flow_matching_loss": 0.01563998917117715, + "grad_norm": 0.17408081889152527, + "learning_rate": 1.1713330151507786e-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": 0.006273146439343691, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.002796279580798, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01534257223829627, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.01563998917117715, + "weighted_rkd_loss": 0.007671286119148135 + }, + { + "epoch": 5.076923076923077, + "flow_matching_loss": 0.018053919449448584, + "grad_norm": 0.18406715989112854, + "learning_rate": 1.161660532525931e-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": 0.006768700829707086, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.002817504294216633, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01635490609332919, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.018053919449448584, + "weighted_rkd_loss": 0.008177453046664595 + }, + { + "epoch": 5.094017094017094, + "flow_matching_loss": 0.015317833703011274, + "grad_norm": 0.1968030482530594, + "learning_rate": 1.1520028499654584e-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": 0.00502970430534333, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.002209481818135828, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01226889044046402, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.015317833703011274, + "weighted_rkd_loss": 0.00613444522023201 + }, + { + "epoch": 5.111111111111111, + "flow_matching_loss": 0.016654190700501205, + "grad_norm": 0.1656157225370407, + "learning_rate": 1.1423603899276833e-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": 0.005547607317566872, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0024255677242763342, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.013520782347768545, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.016654190700501205, + "weighted_rkd_loss": 0.006760391173884273 + }, + { + "epoch": 5.128205128205128, + "flow_matching_loss": 0.018000909499824046, + "grad_norm": 0.1773122400045395, + "learning_rate": 1.1327335742050442e-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": 0.006210618140175939, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.002675327961333096, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01509656421840191, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.018000909499824046, + "weighted_rkd_loss": 0.007548282109200955 + }, + { + "epoch": 5.145299145299146, + "flow_matching_loss": 0.017652605939656497, + "grad_norm": 0.19575658440589905, + "learning_rate": 1.1231228239056467e-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": 0.0056807628832757475, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0025271081482060253, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.013888633996248245, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.017652605939656497, + "weighted_rkd_loss": 0.006944316998124122 + }, + { + "epoch": 5.162393162393163, + "flow_matching_loss": 0.01770560061559081, + "grad_norm": 0.15567773580551147, + "learning_rate": 1.113528559434845e-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": 0.005423327488824725, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0022221908904612063, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.013068845868110657, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.01770560061559081, + "weighted_rkd_loss": 0.0065344229340553285 + }, + { + "epoch": 5.17948717948718, + "flow_matching_loss": 0.01779159056022763, + "grad_norm": 0.14029021561145782, + "learning_rate": 1.1039512004768476e-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": 0.005052718450315297, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.002226500539109111, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.012331937439739704, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.01779159056022763, + "weighted_rkd_loss": 0.006165968719869852 + }, + { + "epoch": 5.196581196581197, + "flow_matching_loss": 0.01728218114003539, + "grad_norm": 0.15812712907791138, + "learning_rate": 1.0943911659763612e-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": 0.005331483390182257, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.002207448345143348, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.012870415206998586, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.01728218114003539, + "weighted_rkd_loss": 0.006435207603499293 + }, + { + "epoch": 5.213675213675214, + "flow_matching_loss": 0.017938074190169573, + "grad_norm": 0.1635051965713501, + "learning_rate": 1.084848874120266e-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": 0.005749216396361589, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.002487788395956159, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.013986221328377724, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.017938074190169573, + "weighted_rkd_loss": 0.006993110664188862 + }, + { + "epoch": 5.230769230769231, + "flow_matching_loss": 0.01776145091280341, + "grad_norm": 0.21638160943984985, + "learning_rate": 1.0753247423193196e-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": 0.0055127232801169155, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0022512326017022134, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01327667897567153, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.01776145091280341, + "weighted_rkd_loss": 0.006638339487835765 + }, + { + "epoch": 5.247863247863248, + "flow_matching_loss": 0.016698889154940843, + "grad_norm": 0.1955612450838089, + "learning_rate": 1.0658191871899012e-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": 0.005643902649171651, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0020833177142776547, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01337112309411168, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.016698889154940843, + "weighted_rkd_loss": 0.00668556154705584 + }, + { + "epoch": 5.264957264957265, + "flow_matching_loss": 0.018136249482631685, + "grad_norm": 0.18112897872924805, + "learning_rate": 1.0563326245357852e-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": 0.006050816411152482, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0026447852607816456, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.014746417943388223, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.018136249482631685, + "weighted_rkd_loss": 0.0073732089716941115 + }, + { + "epoch": 5.282051282051282, + "flow_matching_loss": 0.016763995122164487, + "grad_norm": 0.14852160215377808, + "learning_rate": 1.0468654693299557e-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": 0.006076430296525359, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.00255180976819247, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.014704670384526252, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.016763995122164487, + "weighted_rkd_loss": 0.007352335192263126 + }, + { + "epoch": 5.299145299145299, + "flow_matching_loss": 0.01793050765991211, + "grad_norm": 0.14685483276844025, + "learning_rate": 1.0374181356964501e-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": 0.005478386953473091, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.002295808936469257, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.013252582773566246, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.01793050765991211, + "weighted_rkd_loss": 0.006626291386783123 + }, + { + "epoch": 5.316239316239316, + "flow_matching_loss": 0.016466488875448704, + "grad_norm": 0.18847137689590454, + "learning_rate": 1.0279910368922465e-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": 0.005654708738438785, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.002288639184553176, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.013598056323826314, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.016466488875448704, + "weighted_rkd_loss": 0.006799028161913157 + }, + { + "epoch": 5.333333333333333, + "flow_matching_loss": 0.01684662764891982, + "grad_norm": 0.14454138278961182, + "learning_rate": 1.018584585289186e-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": 0.005896137282252312, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.002443724067416042, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.014235998783260585, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.01684662764891982, + "weighted_rkd_loss": 0.007117999391630292 + }, + { + "epoch": 5.35042735042735, + "flow_matching_loss": 0.019074271619319915, + "grad_norm": 0.2391997128725052, + "learning_rate": 1.0091991923559348e-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": 0.0058216626523062585, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0024093637708574533, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01405268907546997, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.019074271619319915, + "weighted_rkd_loss": 0.007026344537734985 + }, + { + "epoch": 5.367521367521368, + "flow_matching_loss": 0.01673125522211194, + "grad_norm": 0.15762180089950562, + "learning_rate": 9.99835268639984e-06, + "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": 0.00576824105810374, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.002436055976431817, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01397253768518567, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.01673125522211194, + "weighted_rkd_loss": 0.006986268842592835 + }, + { + "epoch": 5.384615384615385, + "flow_matching_loss": 0.015424938220530749, + "grad_norm": 0.18702711164951324, + "learning_rate": 9.904932237496913e-06, + "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": 0.005303484806790948, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0022581320954486728, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.012865101406350731, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.015424938220530749, + "weighted_rkd_loss": 0.006432550703175366 + }, + { + "epoch": 5.401709401709402, + "flow_matching_loss": 0.015545273665338755, + "grad_norm": 0.20115116238594055, + "learning_rate": 9.811734663363659e-06, + "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": 0.005540975974872709, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.002636688796337694, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.013718640618026256, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.015545273665338755, + "weighted_rkd_loss": 0.006859320309013128 + }, + { + "epoch": 5.418803418803419, + "flow_matching_loss": 0.017876550648361446, + "grad_norm": 0.1556462049484253, + "learning_rate": 9.718764040763875e-06, + "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": 0.00610649767331779, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.003013874718453735, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01522686993703246, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.017876550648361446, + "weighted_rkd_loss": 0.00761343496851623 + }, + { + "epoch": 5.435897435897436, + "flow_matching_loss": 0.015284168068319559, + "grad_norm": 0.2237396538257599, + "learning_rate": 9.626024436533777e-06, + "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": 0.005981934443116188, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.00246106336126104, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01442493237555027, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.015284168068319559, + "weighted_rkd_loss": 0.007212466187775135 + }, + { + "epoch": 5.452991452991453, + "flow_matching_loss": 0.01647183932363987, + "grad_norm": 0.22805136442184448, + "learning_rate": 9.533519907404095e-06, + "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": 0.005533198988996446, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.002371192118152976, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.013437590096145868, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.01647183932363987, + "weighted_rkd_loss": 0.006718795048072934 + }, + { + "epoch": 5.47008547008547, + "flow_matching_loss": 0.01914637051522732, + "grad_norm": 0.15872231125831604, + "learning_rate": 9.4412544998226e-06, + "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": 0.006629107845947146, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.002984438475687057, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.016242654342204333, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.01914637051522732, + "weighted_rkd_loss": 0.008121327171102167 + }, + { + "epoch": 5.487179487179487, + "flow_matching_loss": 0.017620038706809282, + "grad_norm": 0.1488192230463028, + "learning_rate": 9.349232249777124e-06, + "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": 0.005304252705536782, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0021423382102511825, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.012750843539834023, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.017620038706809282, + "weighted_rkd_loss": 0.006375421769917011 + }, + { + "epoch": 5.504273504273504, + "flow_matching_loss": 0.020184773299843074, + "grad_norm": 0.1621168553829193, + "learning_rate": 9.257457182618985e-06, + "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": 0.0056879224721342325, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.002303929906338453, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.013679774757474662, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.020184773299843074, + "weighted_rkd_loss": 0.006839887378737331 + }, + { + "epoch": 5.521367521367521, + "flow_matching_loss": 0.016489654313772917, + "grad_norm": 0.17492899298667908, + "learning_rate": 9.165933312886944e-06, + "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": 0.005365319689735771, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.00215536889154464, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.012886008247733116, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.016489654313772917, + "weighted_rkd_loss": 0.006443004123866558 + }, + { + "epoch": 5.538461538461538, + "flow_matching_loss": 0.016707110032439233, + "grad_norm": 0.17003421485424042, + "learning_rate": 9.07466464413155e-06, + "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": 0.005001226905733347, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0022874437272548676, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.012289897538721561, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.016707110032439233, + "weighted_rkd_loss": 0.006144948769360781 + }, + { + "epoch": 5.555555555555555, + "flow_matching_loss": 0.01858956813812256, + "grad_norm": 0.15823964774608612, + "learning_rate": 8.983655168740048e-06, + "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": 0.005134093156084418, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0019241061294451357, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.012192292232066393, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.01858956813812256, + "weighted_rkd_loss": 0.006096146116033196 + }, + { + "epoch": 5.572649572649572, + "flow_matching_loss": 0.01682747919112444, + "grad_norm": 0.19594405591487885, + "learning_rate": 8.892908867761718e-06, + "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": 0.0058005831204354765, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0024930920801125465, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.014094258332625032, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.01682747919112444, + "weighted_rkd_loss": 0.007047129166312516 + }, + { + "epoch": 5.589743589743589, + "flow_matching_loss": 0.01642184555530548, + "grad_norm": 0.17393237352371216, + "learning_rate": 8.802429710733745e-06, + "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": 0.00525958186481148, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.002003428607713431, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.012522592302411794, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.01642184555530548, + "weighted_rkd_loss": 0.006261296151205897 + }, + { + "epoch": 5.6068376068376065, + "flow_matching_loss": 0.015249202214181422, + "grad_norm": 0.15522201359272003, + "learning_rate": 8.71222165550757e-06, + "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": 0.004901192313991487, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.002037578250747174, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.011839963030070066, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.015249202214181422, + "weighted_rkd_loss": 0.005919981515035033 + }, + { + "epoch": 5.6239316239316235, + "flow_matching_loss": 0.015745933819562196, + "grad_norm": 0.14974047243595123, + "learning_rate": 8.622288648075748e-06, + "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": 0.005023935274221003, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0020395138766616584, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01208738461136818, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.015745933819562196, + "weighted_rkd_loss": 0.00604369230568409 + }, + { + "epoch": 5.641025641025641, + "flow_matching_loss": 0.01888891002163291, + "grad_norm": 0.12535054981708527, + "learning_rate": 8.532634622399391e-06, + "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": 0.004384210170246661, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.001768164848908782, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.010536585142835975, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.01888891002163291, + "weighted_rkd_loss": 0.0052682925714179875 + }, + { + "epoch": 5.6581196581196584, + "flow_matching_loss": 0.014870642125606537, + "grad_norm": 0.15545900166034698, + "learning_rate": 8.44326350023601e-06, + "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": 0.0048263138625770806, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0017976223956793548, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.011450250167399645, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.014870642125606537, + "weighted_rkd_loss": 0.005725125083699822 + }, + { + "epoch": 5.6752136752136755, + "flow_matching_loss": 0.01560269109904766, + "grad_norm": 0.16620269417762756, + "learning_rate": 8.354179190968002e-06, + "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": 0.004752189433202147, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0019955436349846424, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.011499922536313534, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.01560269109904766, + "weighted_rkd_loss": 0.005749961268156767 + }, + { + "epoch": 5.6923076923076925, + "flow_matching_loss": 0.018268548417836428, + "grad_norm": 0.19080553948879242, + "learning_rate": 8.265385591431661e-06, + "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": 0.005082469270564616, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0020615262910723686, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.012226464878767728, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.018268548417836428, + "weighted_rkd_loss": 0.006113232439383864 + }, + { + "epoch": 5.7094017094017095, + "flow_matching_loss": 0.016987710259854794, + "grad_norm": 0.176350936293602, + "learning_rate": 8.176886585746671e-06, + "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": 0.005133891617879272, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0020453140372410417, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.012313097342848777, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.016987710259854794, + "weighted_rkd_loss": 0.006156548671424389 + }, + { + "epoch": 5.726495726495727, + "flow_matching_loss": 0.0147684495896101, + "grad_norm": 0.25896912813186646, + "learning_rate": 8.088686045146247e-06, + "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": 0.005305347871035338, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0020323313889093696, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.012643027165904642, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.0147684495896101, + "weighted_rkd_loss": 0.006321513582952321 + }, + { + "epoch": 5.743589743589744, + "flow_matching_loss": 0.016383587382733822, + "grad_norm": 0.16604885458946228, + "learning_rate": 8.00078782780775e-06, + "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": 0.005067420774139464, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0021367019740864634, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01227154340595007, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.016383587382733822, + "weighted_rkd_loss": 0.006135771702975035 + }, + { + "epoch": 5.760683760683761, + "flow_matching_loss": 0.018324275407940148, + "grad_norm": 0.22340555489063263, + "learning_rate": 7.913195778683979e-06, + "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": 0.005241550132632255, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0018835455761291088, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.012366645829752087, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.018324275407940148, + "weighted_rkd_loss": 0.006183322914876044 + }, + { + "epoch": 5.777777777777778, + "flow_matching_loss": 0.0162809194996953, + "grad_norm": 0.17706294357776642, + "learning_rate": 7.825913729334908e-06, + "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": 0.005191397480666637, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0020778504433110355, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.012460645381361246, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.0162809194996953, + "weighted_rkd_loss": 0.006230322690680623 + }, + { + "epoch": 5.794871794871795, + "flow_matching_loss": 0.019178271014243364, + "grad_norm": 0.17093808948993683, + "learning_rate": 7.73894549776013e-06, + "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": 0.005111535452306271, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.002210506598930806, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.012433577328920364, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.019178271014243364, + "weighted_rkd_loss": 0.006216788664460182 + }, + { + "epoch": 5.811965811965812, + "flow_matching_loss": 0.015975886210799217, + "grad_norm": 0.13705652952194214, + "learning_rate": 7.65229488823184e-06, + "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": 0.006296888506039977, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.002588064025621861, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01518184095621109, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.015975886210799217, + "weighted_rkd_loss": 0.007590920478105545 + }, + { + "epoch": 5.829059829059829, + "flow_matching_loss": 0.017226657923310996, + "grad_norm": 0.1626628041267395, + "learning_rate": 7.565965691128396e-06, + "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": 0.005771750584244728, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0024738962412811814, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.014017397444695235, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.017226657923310996, + "weighted_rkd_loss": 0.007008698722347617 + }, + { + "epoch": 5.846153846153846, + "flow_matching_loss": 0.01667610052973032, + "grad_norm": 0.18348802626132965, + "learning_rate": 7.47996168276856e-06, + "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": 0.005051493598148227, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0018816259689629078, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.011984613258391619, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.01667610052973032, + "weighted_rkd_loss": 0.005992306629195809 + }, + { + "epoch": 5.863247863247864, + "flow_matching_loss": 0.01668683849275112, + "grad_norm": 0.17090077698230743, + "learning_rate": 7.394286625246258e-06, + "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": 0.005260160844773054, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0023645029636099935, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.012884824723005294, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.01668683849275112, + "weighted_rkd_loss": 0.006442412361502647 + }, + { + "epoch": 5.880341880341881, + "flow_matching_loss": 0.015944896265864372, + "grad_norm": 0.12532638013362885, + "learning_rate": 7.308944266266063e-06, + "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": 0.004895260930061341, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0019575991085730494, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.011748120933771134, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.015944896265864372, + "weighted_rkd_loss": 0.005874060466885567 + }, + { + "epoch": 5.897435897435898, + "flow_matching_loss": 0.017557265516370536, + "grad_norm": 0.14915937185287476, + "learning_rate": 7.223938338979233e-06, + "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": 0.0045055771013721825, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.001900498615577817, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.010911652911454439, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.017557265516370536, + "weighted_rkd_loss": 0.005455826455727219 + }, + { + "epoch": 5.914529914529915, + "flow_matching_loss": 0.018306993320584296, + "grad_norm": 0.15883104503154755, + "learning_rate": 7.139272561820397e-06, + "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": 0.004605731996707618, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0018143730005249381, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.011025837063789368, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.018306993320584296, + "weighted_rkd_loss": 0.005512918531894684 + }, + { + "epoch": 5.931623931623932, + "flow_matching_loss": 0.015725374035537244, + "grad_norm": 0.1372993290424347, + "learning_rate": 7.054950638344936e-06, + "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": 0.00528130738530308, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.002144762233365327, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.012707377038896083, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.015725374035537244, + "weighted_rkd_loss": 0.006353688519448042 + }, + { + "epoch": 5.948717948717949, + "flow_matching_loss": 0.015682957507669924, + "grad_norm": 0.1382264494895935, + "learning_rate": 6.970976257066941e-06, + "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": 0.004367856332100928, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0018506806693039835, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.010586393298581243, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.015682957507669924, + "weighted_rkd_loss": 0.005293196649290621 + }, + { + "epoch": 5.965811965811966, + "flow_matching_loss": 0.017090506479144098, + "grad_norm": 0.16394349932670593, + "learning_rate": 6.887353091297896e-06, + "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": 0.0051696942886337634, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0020588471903465687, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.012398235965520144, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.017090506479144098, + "weighted_rkd_loss": 0.006199117982760072 + }, + { + "epoch": 5.982905982905983, + "flow_matching_loss": 0.01643017726019025, + "grad_norm": 0.174118310213089, + "learning_rate": 6.804084798985965e-06, + "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": 0.004981178464367986, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.002098930941428989, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.012061287928372622, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.01643017726019025, + "weighted_rkd_loss": 0.006030643964186311 + }, + { + "epoch": 6.0, + "flow_matching_loss": 0.01778987664729357, + "grad_norm": 0.193587064743042, + "learning_rate": 6.721175022556009e-06, + "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": 0.00433569997549057, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0017930015805177391, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.010464401543140411, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.01778987664729357, + "weighted_rkd_loss": 0.0052322007715702055 + }, + { + "epoch": 6.017094017094017, + "flow_matching_loss": 0.01708282008767128, + "grad_norm": 0.16264356672763824, + "learning_rate": 6.638627388750245e-06, + "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": 0.004351393855176866, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0019219616777263581, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.010624749260023237, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.01708282008767128, + "weighted_rkd_loss": 0.005312374630011618 + }, + { + "epoch": 6.034188034188034, + "flow_matching_loss": 0.017007058300077916, + "grad_norm": 0.13111697137355804, + "learning_rate": 6.556445508469583e-06, + "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": 0.005663610715419054, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.002386491384822875, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.013713712897151708, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.017007058300077916, + "weighted_rkd_loss": 0.006856856448575854 + }, + { + "epoch": 6.051282051282051, + "flow_matching_loss": 0.014417500142008067, + "grad_norm": 0.1420252025127411, + "learning_rate": 6.474632976615708e-06, + "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": 0.004468375444412231, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.001813959248829633, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.010750709939748049, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.014417500142008067, + "weighted_rkd_loss": 0.0053753549698740246 + }, + { + "epoch": 6.068376068376068, + "flow_matching_loss": 0.01604814138263464, + "grad_norm": 0.17824363708496094, + "learning_rate": 6.393193371933791e-06, + "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": 0.004844561335630715, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0020488401991315187, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01173796271905303, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.01604814138263464, + "weighted_rkd_loss": 0.005868981359526515 + }, + { + "epoch": 6.085470085470085, + "flow_matching_loss": 0.01477200910449028, + "grad_norm": 0.17427536845207214, + "learning_rate": 6.312130256855977e-06, + "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": 0.0044063873356208205, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0020096577471122147, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.010822432395070792, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.01477200910449028, + "weighted_rkd_loss": 0.005411216197535396 + }, + { + "epoch": 6.102564102564102, + "flow_matching_loss": 0.0156622763723135, + "grad_norm": 0.13360509276390076, + "learning_rate": 6.23144717734552e-06, + "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": 0.004791768454015255, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0019348709029145539, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.011518407706171274, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.0156622763723135, + "weighted_rkd_loss": 0.005759203853085637 + }, + { + "epoch": 6.119658119658119, + "flow_matching_loss": 0.016918635834008454, + "grad_norm": 0.15092743933200836, + "learning_rate": 6.151147662741695e-06, + "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": 0.004752983641810715, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0020754648372530936, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.011581432167440653, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.016918635834008454, + "weighted_rkd_loss": 0.005790716083720326 + }, + { + "epoch": 6.136752136752137, + "flow_matching_loss": 0.016774132382124662, + "grad_norm": 0.13906005024909973, + "learning_rate": 6.071235225605415e-06, + "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": 0.004966419702395797, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0019617047160863875, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.011894544120877981, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.016774132382124662, + "weighted_rkd_loss": 0.005947272060438991 + }, + { + "epoch": 6.153846153846154, + "flow_matching_loss": 0.015961457509547472, + "grad_norm": 0.1445961594581604, + "learning_rate": 5.991713361565546e-06, + "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": 0.005250311200506985, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.002267113933339715, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.012767736334353685, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.015961457509547472, + "weighted_rkd_loss": 0.006383868167176843 + }, + { + "epoch": 6.170940170940171, + "flow_matching_loss": 0.015231370739638806, + "grad_norm": 0.16246339678764343, + "learning_rate": 5.912585549166051e-06, + "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": 0.00495074198115617, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0018103821785189211, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.011711866315454244, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.015231370739638806, + "weighted_rkd_loss": 0.005855933157727122 + }, + { + "epoch": 6.188034188034188, + "flow_matching_loss": 0.01628443393856287, + "grad_norm": 0.16113334894180298, + "learning_rate": 5.833855249713777e-06, + "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": 0.004770028358325362, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0018342768074944616, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.011374333407729864, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.01628443393856287, + "weighted_rkd_loss": 0.005687166703864932 + }, + { + "epoch": 6.205128205128205, + "flow_matching_loss": 0.01632646331563592, + "grad_norm": 0.15477195382118225, + "learning_rate": 5.7555259071270645e-06, + "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": 0.004430649639107287, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.001690405874978751, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.010551705118268728, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.01632646331563592, + "weighted_rkd_loss": 0.005275852559134364 + }, + { + "epoch": 6.222222222222222, + "flow_matching_loss": 0.014321347884833812, + "grad_norm": 0.16369390487670898, + "learning_rate": 5.677600947785117e-06, + "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": 0.005112451920285821, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.002011882374063134, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01223678607493639, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.014321347884833812, + "weighted_rkd_loss": 0.006118393037468195 + }, + { + "epoch": 6.239316239316239, + "flow_matching_loss": 0.015911564882844686, + "grad_norm": 0.1967838704586029, + "learning_rate": 5.600083780378099e-06, + "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": 0.004792532348074019, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0018519766395911573, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.011437041219323874, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.015911564882844686, + "weighted_rkd_loss": 0.005718520609661937 + }, + { + "epoch": 6.256410256410256, + "flow_matching_loss": 0.015266428980976344, + "grad_norm": 0.23702868819236755, + "learning_rate": 5.522977795758038e-06, + "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": 0.004600101965479553, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0017280104337260127, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01092821448110044, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.015266428980976344, + "weighted_rkd_loss": 0.00546410724055022 + }, + { + "epoch": 6.273504273504273, + "flow_matching_loss": 0.016469865664839746, + "grad_norm": 0.17626021802425385, + "learning_rate": 5.4462863667904805e-06, + "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": 0.005159393814392388, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0018100912333466113, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.012128879036754369, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.016469865664839746, + "weighted_rkd_loss": 0.0060644395183771845 + }, + { + "epoch": 6.2905982905982905, + "flow_matching_loss": 0.01662844158709049, + "grad_norm": 0.2012549340724945, + "learning_rate": 5.370012848206983e-06, + "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": 0.004864910733886063, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0020002486533485355, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.011730070039629937, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.01662844158709049, + "weighted_rkd_loss": 0.0058650350198149685 + }, + { + "epoch": 6.3076923076923075, + "flow_matching_loss": 0.015870072040706874, + "grad_norm": 0.14711031317710876, + "learning_rate": 5.29416057645834e-06, + "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": 0.004448941955342889, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0017441964941099287, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.010642080334946513, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.015870072040706874, + "weighted_rkd_loss": 0.005321040167473257 + }, + { + "epoch": 6.3247863247863245, + "flow_matching_loss": 0.016839191410690547, + "grad_norm": 0.1470308154821396, + "learning_rate": 5.218732869568635e-06, + "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": 0.004482673481106758, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.00199831958161667, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.010963666439056396, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.016839191410690547, + "weighted_rkd_loss": 0.005481833219528198 + }, + { + "epoch": 6.3418803418803416, + "flow_matching_loss": 0.015062647126615047, + "grad_norm": 0.15824107825756073, + "learning_rate": 5.14373302699013e-06, + "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": 0.005026099318638444, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0021568586002103985, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.012209057342261076, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.015062647126615047, + "weighted_rkd_loss": 0.006104528671130538 + }, + { + "epoch": 6.358974358974359, + "flow_matching_loss": 0.016755939926952124, + "grad_norm": 0.17414061725139618, + "learning_rate": 5.069164329458894e-06, + "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": 0.004771853005513549, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0017904155771248042, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.011334121599793434, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.016755939926952124, + "weighted_rkd_loss": 0.005667060799896717 + }, + { + "epoch": 6.3760683760683765, + "flow_matching_loss": 0.017453379463404417, + "grad_norm": 0.18111607432365417, + "learning_rate": 4.995030038851342e-06, + "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": 0.004714675666764379, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0018279231037013233, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.011257274448871613, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.017453379463404417, + "weighted_rkd_loss": 0.0056286372244358064 + }, + { + "epoch": 6.3931623931623935, + "flow_matching_loss": 0.01427514124661684, + "grad_norm": 0.12923075258731842, + "learning_rate": 4.9213333980415e-06, + "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": 0.004160228231921792, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0016008496284484863, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.009921306138858199, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.01427514124661684, + "weighted_rkd_loss": 0.004960653069429099 + }, + { + "epoch": 6.410256410256411, + "flow_matching_loss": 0.015912225097417833, + "grad_norm": 0.14404980838298798, + "learning_rate": 4.848077630759175e-06, + "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": 0.005185905285179615, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0022120023844763637, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.0125838128849864, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.015912225097417833, + "weighted_rkd_loss": 0.0062919064424932 + }, + { + "epoch": 6.427350427350428, + "flow_matching_loss": 0.016386830247938633, + "grad_norm": 0.15200337767601013, + "learning_rate": 4.775265941448951e-06, + "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": 0.0058306468650698665, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0022147785755805673, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01387607241049409, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.016386830247938633, + "weighted_rkd_loss": 0.006938036205247045 + }, + { + "epoch": 6.444444444444445, + "flow_matching_loss": 0.015980614442378284, + "grad_norm": 0.1789533942937851, + "learning_rate": 4.702901515129989e-06, + "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": 0.004540768498554826, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.002051534247584641, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.011133071221411227, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.015980614442378284, + "weighted_rkd_loss": 0.005566535610705614 + }, + { + "epoch": 6.461538461538462, + "flow_matching_loss": 0.016609941609203815, + "grad_norm": 0.13359495997428894, + "learning_rate": 4.630987517256734e-06, + "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": 0.00556792204733938, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0021242573857307434, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.013260101620107888, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.016609941609203815, + "weighted_rkd_loss": 0.006630050810053944 + }, + { + "epoch": 6.478632478632479, + "flow_matching_loss": 0.01568589322268963, + "grad_norm": 0.21237364411354065, + "learning_rate": 4.559527093580412e-06, + "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": 0.005020906263962388, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0020745873218402265, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.012116399919614195, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.01568589322268963, + "weighted_rkd_loss": 0.006058199959807098 + }, + { + "epoch": 6.495726495726496, + "flow_matching_loss": 0.015495562460273504, + "grad_norm": 0.1483098566532135, + "learning_rate": 4.488523370011456e-06, + "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": 0.004437587270513177, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.001733652234543115, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.010608826857060194, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.015495562460273504, + "weighted_rkd_loss": 0.005304413428530097 + }, + { + "epoch": 6.512820512820513, + "flow_matching_loss": 0.014806710742413997, + "grad_norm": 0.1273726522922516, + "learning_rate": 4.417979452482766e-06, + "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": 0.004741726303473115, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0019353451789356768, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.011418797774240375, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.014806710742413997, + "weighted_rkd_loss": 0.005709398887120187 + }, + { + "epoch": 6.52991452991453, + "flow_matching_loss": 0.016286912932991983, + "grad_norm": 0.14928042888641357, + "learning_rate": 4.347898426813811e-06, + "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": 0.004060622188262641, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0015757541987113655, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.00969699863344431, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.016286912932991983, + "weighted_rkd_loss": 0.004848499316722155 + }, + { + "epoch": 6.547008547008547, + "flow_matching_loss": 0.01606764690950513, + "grad_norm": 0.15828359127044678, + "learning_rate": 4.278283358575694e-06, + "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": 0.00493182628415525, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0019138668780215085, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.011777519434690475, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.01606764690950513, + "weighted_rkd_loss": 0.005888759717345237 + }, + { + "epoch": 6.564102564102564, + "flow_matching_loss": 0.01618655240163207, + "grad_norm": 0.22989870607852936, + "learning_rate": 4.2091372929570145e-06, + "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": 0.005280993832275272, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.002411555079743266, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01297354269772768, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.01618655240163207, + "weighted_rkd_loss": 0.00648677134886384 + }, + { + "epoch": 6.581196581196581, + "flow_matching_loss": 0.01933974390849471, + "grad_norm": 0.16607911884784698, + "learning_rate": 4.1404632546306934e-06, + "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": 0.00440069807227701, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0018826108775101603, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01068400703370571, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.01933974390849471, + "weighted_rkd_loss": 0.005342003516852855 + }, + { + "epoch": 6.598290598290598, + "flow_matching_loss": 0.016087422892451287, + "grad_norm": 0.17650851607322693, + "learning_rate": 4.0722642476216286e-06, + "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": 0.004229920823127032, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0016205348074436188, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01008037650026381, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.016087422892451287, + "weighted_rkd_loss": 0.005040188250131905 + }, + { + "epoch": 6.615384615384615, + "flow_matching_loss": 0.015177225507795811, + "grad_norm": 0.1870707869529724, + "learning_rate": 4.00454325517533e-06, + "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": 0.005106892134062946, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.002013124746736139, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.012226908840239048, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.015177225507795811, + "weighted_rkd_loss": 0.006113454420119524 + }, + { + "epoch": 6.632478632478632, + "flow_matching_loss": 0.015384182892739772, + "grad_norm": 0.165043905377388, + "learning_rate": 3.937303239627393e-06, + "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": 0.0036144151352345945, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.001538443169556558, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.008767273370176553, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.015384182892739772, + "weighted_rkd_loss": 0.004383636685088277 + }, + { + "epoch": 6.64957264957265, + "flow_matching_loss": 0.016996524203568696, + "grad_norm": 0.2328128218650818, + "learning_rate": 3.870547142273918e-06, + "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": 0.004557771165855229, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0020668866171035917, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.011182428896427154, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.016996524203568696, + "weighted_rkd_loss": 0.005591214448213577 + }, + { + "epoch": 6.666666666666667, + "flow_matching_loss": 0.016605217289179564, + "grad_norm": 0.18834178149700165, + "learning_rate": 3.8042778832428736e-06, + "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": 0.004030448896810412, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0015587454079650342, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.009619643166661262, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.016605217289179564, + "weighted_rkd_loss": 0.004809821583330631 + }, + { + "epoch": 6.683760683760684, + "flow_matching_loss": 0.019808900728821754, + "grad_norm": 0.17286179959774017, + "learning_rate": 3.7384983613663195e-06, + "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": 0.004534973669797182, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0016975721111521125, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.010767519474029541, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.019808900728821754, + "weighted_rkd_loss": 0.0053837597370147705 + }, + { + "epoch": 6.700854700854701, + "flow_matching_loss": 0.017337767593562602, + "grad_norm": 0.16282841563224792, + "learning_rate": 3.6732114540536503e-06, + "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": 0.0041676078690215945, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0015985081496182828, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.009933723881840707, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.017337767593562602, + "weighted_rkd_loss": 0.004966861940920353 + }, + { + "epoch": 6.717948717948718, + "flow_matching_loss": 0.016870119236409666, + "grad_norm": 0.15538223087787628, + "learning_rate": 3.608420017165683e-06, + "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": 0.005332878767512738, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0020503941224887965, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.012716151541098952, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.016870119236409666, + "weighted_rkd_loss": 0.006358075770549476 + }, + { + "epoch": 6.735042735042735, + "flow_matching_loss": 0.017999943252652882, + "grad_norm": 0.1980685442686081, + "learning_rate": 3.5441268848897664e-06, + "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": 0.00407553631812334, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0015552254626527429, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.009706298168748618, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.017999943252652882, + "weighted_rkd_loss": 0.004853149084374309 + }, + { + "epoch": 6.752136752136752, + "flow_matching_loss": 0.014841801393777132, + "grad_norm": 0.14662159979343414, + "learning_rate": 3.480334869615792e-06, + "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": 0.0043201320571824905, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.001651768758893013, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.010292032966390252, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.014841801393777132, + "weighted_rkd_loss": 0.005146016483195126 + }, + { + "epoch": 6.769230769230769, + "flow_matching_loss": 0.014047346450388432, + "grad_norm": 0.1611618548631668, + "learning_rate": 3.417046761813155e-06, + "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": 0.004797124327160418, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0017810778925195336, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.011375326476991177, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.014047346450388432, + "weighted_rkd_loss": 0.005687663238495589 + }, + { + "epoch": 6.786324786324786, + "flow_matching_loss": 0.016620884370058774, + "grad_norm": 0.2051396518945694, + "learning_rate": 3.3542653299087263e-06, + "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": 0.004709043470211327, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0018011134234257042, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.011219200305640697, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.016620884370058774, + "weighted_rkd_loss": 0.0056096001528203486 + }, + { + "epoch": 6.803418803418803, + "flow_matching_loss": 0.013227695412933827, + "grad_norm": 0.1768273413181305, + "learning_rate": 3.2919933201657117e-06, + "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": 0.004239079589024186, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0019575843820348383, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.010435743583366274, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.013227695412933827, + "weighted_rkd_loss": 0.005217871791683137 + }, + { + "epoch": 6.82051282051282, + "flow_matching_loss": 0.016534697357565166, + "grad_norm": 0.18357117474079132, + "learning_rate": 3.2302334565635584e-06, + "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": 0.004425921780057252, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0017305678804405034, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.010582411382347345, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.016534697357565166, + "weighted_rkd_loss": 0.005291205691173672 + }, + { + "epoch": 6.837606837606837, + "flow_matching_loss": 0.014859104063361883, + "grad_norm": 0.21632759273052216, + "learning_rate": 3.1689884406787674e-06, + "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": 0.0038600806845352055, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0016228868742473423, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.009343048324808478, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.014859104063361883, + "weighted_rkd_loss": 0.004671524162404239 + }, + { + "epoch": 6.854700854700854, + "flow_matching_loss": 0.017187664564698935, + "grad_norm": 0.13118460774421692, + "learning_rate": 3.1082609515667424e-06, + "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": 0.004225519369356334, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0017153617111034692, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.010166400438174605, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.017187664564698935, + "weighted_rkd_loss": 0.0050832002190873025 + }, + { + "epoch": 6.871794871794872, + "flow_matching_loss": 0.016087855119258164, + "grad_norm": 0.194352388381958, + "learning_rate": 3.048053645644594e-06, + "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": 0.004914767970331013, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0019320292049087584, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.011761565040796996, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.016087855119258164, + "weighted_rkd_loss": 0.005880782520398498 + }, + { + "epoch": 6.888888888888889, + "flow_matching_loss": 0.015333417244255543, + "grad_norm": 0.15082167088985443, + "learning_rate": 2.988369156574916e-06, + "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": 0.003941808850504458, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.001569393661338836, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.009453011397272348, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.015333417244255543, + "weighted_rkd_loss": 0.004726505698636174 + }, + { + "epoch": 6.905982905982906, + "flow_matching_loss": 0.016069279052317142, + "grad_norm": 0.13590292632579803, + "learning_rate": 2.9292100951506225e-06, + "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": 0.004573615314438939, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0018226096872240305, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01096984026953578, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.016069279052317142, + "weighted_rkd_loss": 0.00548492013476789 + }, + { + "epoch": 6.923076923076923, + "flow_matching_loss": 0.015894505102187394, + "grad_norm": 0.16080982983112335, + "learning_rate": 2.8705790491807033e-06, + "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": 0.004575653979554772, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0017367414198815822, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01088804923929274, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.015894505102187394, + "weighted_rkd_loss": 0.00544402461964637 + }, + { + "epoch": 6.94017094017094, + "flow_matching_loss": 0.013970594946295023, + "grad_norm": 0.16656595468521118, + "learning_rate": 2.8124785833770433e-06, + "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": 0.004196502268314362, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0016817040625028313, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.010074708424508572, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.013970594946295023, + "weighted_rkd_loss": 0.005037354212254286 + }, + { + "epoch": 6.957264957264957, + "flow_matching_loss": 0.017886763066053392, + "grad_norm": 0.15579462051391602, + "learning_rate": 2.754911239242241e-06, + "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": 0.004205319180618972, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0021213955304119738, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.010532033909112216, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.017886763066053392, + "weighted_rkd_loss": 0.005266016954556108 + }, + { + "epoch": 6.9743589743589745, + "flow_matching_loss": 0.01466821664944291, + "grad_norm": 0.1741836667060852, + "learning_rate": 2.6978795349584204e-06, + "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": 0.0048158891731873155, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.001956576685188338, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.011588354967534542, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.01466821664944291, + "weighted_rkd_loss": 0.005794177483767271 + }, + { + "epoch": 6.9914529914529915, + "flow_matching_loss": 0.016218194179236887, + "grad_norm": 0.15357689559459686, + "learning_rate": 2.641385965277088e-06, + "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": 0.004819849249906838, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.001966080185957253, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.011605778662487864, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.016218194179236887, + "weighted_rkd_loss": 0.005802889331243932 + }, + { + "epoch": 7.0085470085470085, + "flow_matching_loss": 0.01698446348309517, + "grad_norm": 0.2071133255958557, + "learning_rate": 2.5854330014099923e-06, + "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": 0.004636326502077281, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0016064835479483008, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01087913648225367, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.01698446348309517, + "weighted_rkd_loss": 0.005439568241126835 + }, + { + "epoch": 7.0256410256410255, + "flow_matching_loss": 0.016615198738873006, + "grad_norm": 0.2097112238407135, + "learning_rate": 2.5300230909210298e-06, + "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": 0.004680479527451098, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.001894661597907543, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.011255620792508126, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.016615198738873006, + "weighted_rkd_loss": 0.005627810396254063 + }, + { + "epoch": 7.042735042735043, + "flow_matching_loss": 0.014118788111954928, + "grad_norm": 0.1714528203010559, + "learning_rate": 2.4751586576191983e-06, + "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": 0.004255960811860859, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0015853908960707485, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.010097312461584806, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.014118788111954928, + "weighted_rkd_loss": 0.005048656230792403 + }, + { + "epoch": 7.05982905982906, + "flow_matching_loss": 0.01413761992007494, + "grad_norm": 0.14934183657169342, + "learning_rate": 2.420842101452538e-06, + "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": 0.004364838264882565, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0020897478796541693, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.010819424223154784, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.01413761992007494, + "weighted_rkd_loss": 0.005409712111577392 + }, + { + "epoch": 7.076923076923077, + "flow_matching_loss": 0.014027675800025462, + "grad_norm": 0.17164747416973114, + "learning_rate": 2.3670757984031816e-06, + "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": 0.0037490071495994925, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0015432772459462285, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.009041291614994406, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.014027675800025462, + "weighted_rkd_loss": 0.004520645807497203 + }, + { + "epoch": 7.094017094017094, + "flow_matching_loss": 0.013278034422546625, + "grad_norm": 0.12479699403047562, + "learning_rate": 2.313862100383404e-06, + "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": 0.0036123374244198202, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0014923112001270055, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.008716986002400517, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.013278034422546625, + "weighted_rkd_loss": 0.004358493001200258 + }, + { + "epoch": 7.111111111111111, + "flow_matching_loss": 0.015609122812747955, + "grad_norm": 0.11537393182516098, + "learning_rate": 2.2612033351327406e-06, + "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": 0.0045773436082527045, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.002064816007623449, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.011219503311440348, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.015609122812747955, + "weighted_rkd_loss": 0.005609751655720174 + }, + { + "epoch": 7.128205128205128, + "flow_matching_loss": 0.014534916821867227, + "grad_norm": 0.1255650818347931, + "learning_rate": 2.2091018061161788e-06, + "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": 0.003910883702337742, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0016804726677946747, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.009502240177243947, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.014534916821867227, + "weighted_rkd_loss": 0.004751120088621974 + }, + { + "epoch": 7.145299145299146, + "flow_matching_loss": 0.016385517176240684, + "grad_norm": 0.16010573506355286, + "learning_rate": 2.15755979242338e-06, + "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": 0.004504839330911637, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.001814916729927063, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.010824595438316464, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.016385517176240684, + "weighted_rkd_loss": 0.005412297719158232 + }, + { + "epoch": 7.162393162393163, + "flow_matching_loss": 0.01727302558720112, + "grad_norm": 0.16287177801132202, + "learning_rate": 2.106579548669e-06, + "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": 0.003930654772557318, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0013079404830932616, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.009169249981641769, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.01727302558720112, + "weighted_rkd_loss": 0.004584624990820884 + }, + { + "epoch": 7.17948717948718, + "flow_matching_loss": 0.015191889088600873, + "grad_norm": 0.18671517074108124, + "learning_rate": 2.0561633048940477e-06, + "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": 0.004497651406563818, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0019035461591556668, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.010898848949000239, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.015191889088600873, + "weighted_rkd_loss": 0.005449424474500119 + }, + { + "epoch": 7.196581196581197, + "flow_matching_loss": 0.016087721660733224, + "grad_norm": 0.1463557481765747, + "learning_rate": 2.006313266468355e-06, + "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": 0.004812970152124762, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0019313882337883114, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.011557328561320902, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.016087721660733224, + "weighted_rkd_loss": 0.005778664280660451 + }, + { + "epoch": 7.213675213675214, + "flow_matching_loss": 0.016287938598543404, + "grad_norm": 0.1706915646791458, + "learning_rate": 1.957031613994097e-06, + "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": 0.004160731332376599, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0016245927836280316, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.00994605547748506, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.016287938598543404, + "weighted_rkd_loss": 0.00497302773874253 + }, + { + "epoch": 7.230769230769231, + "flow_matching_loss": 0.015654664393514393, + "grad_norm": 0.1440427303314209, + "learning_rate": 1.9083205032103995e-06, + "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": 0.004027574555948377, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.001447567099239677, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.00950271631591022, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.015654664393514393, + "weighted_rkd_loss": 0.00475135815795511 + }, + { + "epoch": 7.247863247863248, + "flow_matching_loss": 0.01606919979676604, + "grad_norm": 0.18371421098709106, + "learning_rate": 1.8601820648990536e-06, + "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": 0.004114628559909761, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0014564043085556477, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.009685661410912871, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.01606919979676604, + "weighted_rkd_loss": 0.004842830705456436 + }, + { + "epoch": 7.264957264957265, + "flow_matching_loss": 0.014553804695606232, + "grad_norm": 0.1450345814228058, + "learning_rate": 1.8126184047912936e-06, + "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": 0.0038839376764371993, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0014871912775561214, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.009255066746845841, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.014553804695606232, + "weighted_rkd_loss": 0.004627533373422921 + }, + { + "epoch": 7.282051282051282, + "flow_matching_loss": 0.013839024305343627, + "grad_norm": 0.22128815948963165, + "learning_rate": 1.7656316034757024e-06, + "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": 0.004872125829569996, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.002003580390010029, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.0117478318978101, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.013839024305343627, + "weighted_rkd_loss": 0.00587391594890505 + }, + { + "epoch": 7.299145299145299, + "flow_matching_loss": 0.016494604852050544, + "grad_norm": 0.14832022786140442, + "learning_rate": 1.7192237163071816e-06, + "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": 0.003961826534941792, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0015516869083512575, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.00947534004226327, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.016494604852050544, + "weighted_rkd_loss": 0.004737670021131635 + }, + { + "epoch": 7.316239316239316, + "flow_matching_loss": 0.015676535665988922, + "grad_norm": 0.14860793948173523, + "learning_rate": 1.6733967733170573e-06, + "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": 0.003999082953669131, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.002201343182241544, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.010199508955702185, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.015676535665988922, + "weighted_rkd_loss": 0.005099754477851093 + }, + { + "epoch": 7.333333333333333, + "flow_matching_loss": 0.015473322197794914, + "grad_norm": 0.17381682991981506, + "learning_rate": 1.628152779124279e-06, + "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": 0.004379562707617879, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0016722492931876332, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.010431374609470367, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.015473322197794914, + "weighted_rkd_loss": 0.005215687304735183 + }, + { + "epoch": 7.35042735042735, + "flow_matching_loss": 0.015086284559220075, + "grad_norm": 0.13543272018432617, + "learning_rate": 1.583493712847715e-06, + "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": 0.00412148276809603, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0015148779144510626, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.00975784338079393, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.015086284559220075, + "weighted_rkd_loss": 0.004878921690396965 + }, + { + "epoch": 7.367521367521368, + "flow_matching_loss": 0.01674106726422906, + "grad_norm": 0.1808241754770279, + "learning_rate": 1.5394215280196084e-06, + "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": 0.003748697927221656, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.001673176372423768, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.009170572226867079, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.01674106726422906, + "weighted_rkd_loss": 0.0045852861134335395 + }, + { + "epoch": 7.384615384615385, + "flow_matching_loss": 0.01622048085555434, + "grad_norm": 0.24213089048862457, + "learning_rate": 1.4959381525000904e-06, + "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": 0.004719791142269969, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0019229460158385336, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.011362528428435326, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.01622048085555434, + "weighted_rkd_loss": 0.005681264214217663 + }, + { + "epoch": 7.401709401709402, + "flow_matching_loss": 0.01691405791789293, + "grad_norm": 0.136305570602417, + "learning_rate": 1.453045488392879e-06, + "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": 0.004566896031610668, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0017233178252354265, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.010857109911739826, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.01691405791789293, + "weighted_rkd_loss": 0.005428554955869913 + }, + { + "epoch": 7.418803418803419, + "flow_matching_loss": 0.015587300807237626, + "grad_norm": 0.1919413059949875, + "learning_rate": 1.4107454119620484e-06, + "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": 0.004084520391188562, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0016211720067076385, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.009790212847292423, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.015587300807237626, + "weighted_rkd_loss": 0.004895106423646212 + }, + { + "epoch": 7.435897435897436, + "flow_matching_loss": 0.016039892192929985, + "grad_norm": 0.18631604313850403, + "learning_rate": 1.3690397735499778e-06, + "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": 0.004030122491531074, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0018015784677118062, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.009861823450773955, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.016039892192929985, + "weighted_rkd_loss": 0.0049309117253869775 + }, + { + "epoch": 7.452991452991453, + "flow_matching_loss": 0.017094092350453137, + "grad_norm": 0.1710483431816101, + "learning_rate": 1.327930397496403e-06, + "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": 0.004542760131880641, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.001764396857470274, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.010849917074665427, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.017094092350453137, + "weighted_rkd_loss": 0.005424958537332714 + }, + { + "epoch": 7.47008547008547, + "flow_matching_loss": 0.01391311539337039, + "grad_norm": 0.1865139752626419, + "learning_rate": 1.287419082058604e-06, + "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": 0.003823025431483984, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.001664593192981556, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.009310644119977951, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.01391311539337039, + "weighted_rkd_loss": 0.0046553220599889755 + }, + { + "epoch": 7.487179487179487, + "flow_matching_loss": 0.015736521501094104, + "grad_norm": 0.20026004314422607, + "learning_rate": 1.2475075993327589e-06, + "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": 0.004398590861819684, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0017495504696853458, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.010546732181683182, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.015736521501094104, + "weighted_rkd_loss": 0.005273366090841591 + }, + { + "epoch": 7.504273504273504, + "flow_matching_loss": 0.015191830042749644, + "grad_norm": 0.12072700262069702, + "learning_rate": 1.208197695176413e-06, + "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": 0.004338941141031683, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.00170200303546153, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01037988536991179, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.015191830042749644, + "weighted_rkd_loss": 0.005189942684955895 + }, + { + "epoch": 7.521367521367521, + "flow_matching_loss": 0.014817466028034687, + "grad_norm": 0.13009797036647797, + "learning_rate": 1.1694910891321259e-06, + "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": 0.003838592185638845, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0016448823153041303, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.009322066651657223, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.014817466028034687, + "weighted_rkd_loss": 0.004661033325828612 + }, + { + "epoch": 7.538461538461538, + "flow_matching_loss": 0.017156030517071486, + "grad_norm": 0.18017660081386566, + "learning_rate": 1.1313894743522363e-06, + "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": 0.00444778697565198, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0019361308077350259, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.010831704782322048, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.017156030517071486, + "weighted_rkd_loss": 0.005415852391161024 + }, + { + "epoch": 7.555555555555555, + "flow_matching_loss": 0.015430319868028165, + "grad_norm": 0.13498713076114655, + "learning_rate": 1.0938945175248016e-06, + "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": 0.003500660159625113, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0014626291289459915, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.00846394938416779, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.015430319868028165, + "weighted_rkd_loss": 0.004231974692083895 + }, + { + "epoch": 7.572649572649572, + "flow_matching_loss": 0.016946085449308156, + "grad_norm": 0.14081688225269318, + "learning_rate": 1.0570078588007086e-06, + "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": 0.0043074612505733965, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0018926878925412892, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.010507610347121953, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.016946085449308156, + "weighted_rkd_loss": 0.005253805173560977 + }, + { + "epoch": 7.589743589743589, + "flow_matching_loss": 0.013496050518006087, + "grad_norm": 0.20047232508659363, + "learning_rate": 1.0207311117219009e-06, + "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": 0.003938631736673415, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0014799494645558297, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.00935721299611032, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.013496050518006087, + "weighted_rkd_loss": 0.00467860649805516 + }, + { + "epoch": 7.6068376068376065, + "flow_matching_loss": 0.014154733717441558, + "grad_norm": 0.20217622816562653, + "learning_rate": 9.850658631508198e-07, + "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": 0.0044426189735531805, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0018720550229772925, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.010757293039932846, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.014154733717441558, + "weighted_rkd_loss": 0.005378646519966423 + }, + { + "epoch": 7.6239316239316235, + "flow_matching_loss": 0.01568802334368229, + "grad_norm": 0.13635766506195068, + "learning_rate": 9.500136732009751e-07, + "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": 0.004027505731210112, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0016500390833243729, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.00970505066215992, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.01568802334368229, + "weighted_rkd_loss": 0.00485252533107996 + }, + { + "epoch": 7.641025641025641, + "flow_matching_loss": 0.013338806852698327, + "grad_norm": 0.1933533251285553, + "learning_rate": 9.155760751687054e-07, + "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": 0.004383315378800034, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0017131514148786664, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.010479782242327929, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.013338806852698327, + "weighted_rkd_loss": 0.005239891121163964 + }, + { + "epoch": 7.6581196581196584, + "flow_matching_loss": 0.015097386622801422, + "grad_norm": 0.1867772787809372, + "learning_rate": 8.817545754661155e-07, + "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": 0.004559310991317034, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.001779672270640731, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.010898294346407055, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.015097386622801422, + "weighted_rkd_loss": 0.005449147173203528 + }, + { + "epoch": 7.6752136752136755, + "flow_matching_loss": 0.017503133229911327, + "grad_norm": 0.1295897662639618, + "learning_rate": 8.485506535551668e-07, + "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": 0.0039037192706018685, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0015860553365200758, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.009393493924289942, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.017503133229911327, + "weighted_rkd_loss": 0.004696746962144971 + }, + { + "epoch": 7.6923076923076925, + "flow_matching_loss": 0.012718126643449068, + "grad_norm": 0.16545599699020386, + "learning_rate": 8.159657618829685e-07, + "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": 0.004243917670100928, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0016703173867426812, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.010158152924850583, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.012718126643449068, + "weighted_rkd_loss": 0.005079076462425292 + }, + { + "epoch": 7.7094017094017095, + "flow_matching_loss": 0.017403970565646888, + "grad_norm": 0.18544143438339233, + "learning_rate": 7.840013258182471e-07, + "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": 0.0040193015011027455, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0016501099104061722, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.009688712889328599, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.017403970565646888, + "weighted_rkd_loss": 0.0048443564446642995 + }, + { + "epoch": 7.726495726495727, + "flow_matching_loss": 0.017219954915344716, + "grad_norm": 0.13798153400421143, + "learning_rate": 7.526587435889809e-07, + "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": 0.0035293982829898594, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.001458873733645305, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.00851767030544579, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.017219954915344716, + "weighted_rkd_loss": 0.004258835152722895 + }, + { + "epoch": 7.743589743589744, + "flow_matching_loss": 0.01541247796267271, + "grad_norm": 0.139375701546669, + "learning_rate": 7.219393862212593e-07, + "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": 0.0038150731241330506, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0014316341024823488, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.009061780339106918, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.01541247796267271, + "weighted_rkd_loss": 0.004530890169553459 + }, + { + "epoch": 7.760683760683761, + "flow_matching_loss": 0.015388681180775165, + "grad_norm": 0.13138681650161743, + "learning_rate": 6.918445974792819e-07, + "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": 0.003366332477889955, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.001311596919549629, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.008044261950999498, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.015388681180775165, + "weighted_rkd_loss": 0.004022130975499749 + }, + { + "epoch": 7.777777777777778, + "flow_matching_loss": 0.014622988179326057, + "grad_norm": 0.11516177654266357, + "learning_rate": 6.6237569380661e-07, + "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": 0.0038457690505310895, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0015454158827196806, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.009236953966319562, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.014622988179326057, + "weighted_rkd_loss": 0.004618476983159781 + }, + { + "epoch": 7.794871794871795, + "flow_matching_loss": 0.014921532105654478, + "grad_norm": 0.13934694230556488, + "learning_rate": 6.335339642685495e-07, + "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": 0.0037269066786393525, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0015626251930370926, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.00901643866673112, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.014921532105654478, + "weighted_rkd_loss": 0.00450821933336556 + }, + { + "epoch": 7.811965811965812, + "flow_matching_loss": 0.015170859359204768, + "grad_norm": 0.20881541073322296, + "learning_rate": 6.05320670495787e-07, + "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": 0.00461851479485631, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0017640399921219796, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.011001069517806172, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.015170859359204768, + "weighted_rkd_loss": 0.005500534758903086 + }, + { + "epoch": 7.829059829059829, + "flow_matching_loss": 0.014338784758001566, + "grad_norm": 0.16537579894065857, + "learning_rate": 5.777370466291954e-07, + "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": 0.004272118653170764, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.001494649751111865, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.010038887057453394, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.014338784758001566, + "weighted_rkd_loss": 0.005019443528726697 + }, + { + "epoch": 7.846153846153846, + "flow_matching_loss": 0.013995079416781664, + "grad_norm": 0.14324721693992615, + "learning_rate": 5.507842992658336e-07, + "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": 0.003364446433261037, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.001459231018088758, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.008188123861327768, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.013995079416781664, + "weighted_rkd_loss": 0.004094061930663884 + }, + { + "epoch": 7.863247863247864, + "flow_matching_loss": 0.0174401942640543, + "grad_norm": 0.1871991753578186, + "learning_rate": 5.24463607406192e-07, + "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": 0.00412391189020127, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0018717199098318815, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.010119543736800551, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.0174401942640543, + "weighted_rkd_loss": 0.0050597718684002755 + }, + { + "epoch": 7.880341880341881, + "flow_matching_loss": 0.013484288938343525, + "grad_norm": 0.16377152502536774, + "learning_rate": 4.987761224025944e-07, + "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": 0.0044992591720074415, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0016240559634752572, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.010622574388980866, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.013484288938343525, + "weighted_rkd_loss": 0.005311287194490433 + }, + { + "epoch": 7.897435897435898, + "flow_matching_loss": 0.017789938021451235, + "grad_norm": 0.1471613347530365, + "learning_rate": 4.737229679088556e-07, + "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": 0.004621709953062237, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.00174307432025671, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.010986494272947312, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.017789938021451235, + "weighted_rkd_loss": 0.005493247136473656 + }, + { + "epoch": 7.914529914529915, + "flow_matching_loss": 0.016959631443023683, + "grad_norm": 0.20595283806324005, + "learning_rate": 4.493052398311115e-07, + "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": 0.004121056059375405, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0017009781731758267, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.009943090286105872, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.016959631443023683, + "weighted_rkd_loss": 0.004971545143052936 + }, + { + "epoch": 7.931623931623932, + "flow_matching_loss": 0.014979766122996807, + "grad_norm": 0.16615192592144012, + "learning_rate": 4.2552400627989043e-07, + "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": 0.004145107767544687, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0016354189720004797, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.009925634413957597, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.014979766122996807, + "weighted_rkd_loss": 0.004962817206978798 + }, + { + "epoch": 7.948717948717949, + "flow_matching_loss": 0.015234688110649586, + "grad_norm": 0.14960739016532898, + "learning_rate": 4.023803075233923e-07, + "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": 0.004770138673484326, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0019104938895907254, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.011450771149247885, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.015234688110649586, + "weighted_rkd_loss": 0.0057253855746239425 + }, + { + "epoch": 7.965811965811966, + "flow_matching_loss": 0.015797104593366383, + "grad_norm": 0.17209915816783905, + "learning_rate": 3.7987515594196997e-07, + "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": 0.0044247575104236605, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0016803916427306832, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.010529906675219535, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.015797104593366383, + "weighted_rkd_loss": 0.005264953337609768 + }, + { + "epoch": 7.982905982905983, + "flow_matching_loss": 0.015738171897828578, + "grad_norm": 0.13701865077018738, + "learning_rate": 3.580095359838648e-07, + "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": 0.004525296087376773, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0018712702090851962, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.010921862442046404, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.015738171897828578, + "weighted_rkd_loss": 0.005460931221023202 + }, + { + "epoch": 8.0, + "flow_matching_loss": 0.015426522772759198, + "grad_norm": 0.14993874728679657, + "learning_rate": 3.3678440412212576e-07, + "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": 0.0044903341913595796, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0018846069695428013, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.010865275282412767, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.015426522772759198, + "weighted_rkd_loss": 0.005432637641206384 + }, + { + "epoch": 8.017094017094017, + "flow_matching_loss": 0.016985416971147062, + "grad_norm": 0.17564892768859863, + "learning_rate": 3.162006888127833e-07, + "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": 0.00391091569326818, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0014062735368497669, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.00922810472548008, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.016985416971147062, + "weighted_rkd_loss": 0.00461405236274004 + }, + { + "epoch": 8.034188034188034, + "flow_matching_loss": 0.017536366637796164, + "grad_norm": 0.16656459867954254, + "learning_rate": 2.9625929045422327e-07, + "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": 0.0039096885360777375, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.001401591202011332, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.009220968279987574, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.017536366637796164, + "weighted_rkd_loss": 0.004610484139993787 + }, + { + "epoch": 8.051282051282051, + "flow_matching_loss": 0.016558579914271832, + "grad_norm": 0.16844052076339722, + "learning_rate": 2.7696108134781186e-07, + "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": 0.004956786707043648, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0017663251608610153, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.011679898528382183, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.016558579914271832, + "weighted_rkd_loss": 0.005839949264191091 + }, + { + "epoch": 8.068376068376068, + "flow_matching_loss": 0.016213190834969284, + "grad_norm": 0.16253527998924255, + "learning_rate": 2.583069056597376e-07, + "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": 0.004380698874592781, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0020641373936086892, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.010825535003095865, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.016213190834969284, + "weighted_rkd_loss": 0.005412767501547932 + }, + { + "epoch": 8.085470085470085, + "flow_matching_loss": 0.013580971863120794, + "grad_norm": 0.12660358846187592, + "learning_rate": 2.4029757938407116e-07, + "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": 0.004097833367995918, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.001680993870832026, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.009876660583540798, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.013580971863120794, + "weighted_rkd_loss": 0.004938330291770399 + }, + { + "epoch": 8.102564102564102, + "flow_matching_loss": 0.01842335294932127, + "grad_norm": 0.18897368013858795, + "learning_rate": 2.229338903070871e-07, + "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": 0.004228601953946054, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0018422544933855533, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.010299458540976048, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.01842335294932127, + "weighted_rkd_loss": 0.005149729270488024 + }, + { + "epoch": 8.11965811965812, + "flow_matching_loss": 0.01590350326150656, + "grad_norm": 0.20238161087036133, + "learning_rate": 2.0621659797279645e-07, + "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": 0.004119533277116716, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0015901223989203573, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.009829188929870725, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.01590350326150656, + "weighted_rkd_loss": 0.004914594464935362 + }, + { + "epoch": 8.136752136752136, + "flow_matching_loss": 0.015774535667151212, + "grad_norm": 0.12638553977012634, + "learning_rate": 1.901464336497233e-07, + "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": 0.003907700185663998, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.001608267182018608, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.009423667518422007, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.015774535667151212, + "weighted_rkd_loss": 0.004711833759211003 + }, + { + "epoch": 8.153846153846153, + "flow_matching_loss": 0.016548135597258807, + "grad_norm": 0.16260124742984772, + "learning_rate": 1.7472410029891527e-07, + "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": 0.0036436608759686353, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.00162093035178259, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.00890825199894607, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.016548135597258807, + "weighted_rkd_loss": 0.004454125999473035 + }, + { + "epoch": 8.17094017094017, + "flow_matching_loss": 0.015792224556207657, + "grad_norm": 0.1684548705816269, + "learning_rate": 1.5995027254319173e-07, + "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": 0.003674332913942635, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.001533985964488238, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.008882651710882783, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.015792224556207657, + "weighted_rkd_loss": 0.004441325855441392 + }, + { + "epoch": 8.188034188034187, + "flow_matching_loss": 0.015294800698757171, + "grad_norm": 0.1381729245185852, + "learning_rate": 1.4582559663764372e-07, + "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": 0.004142222669906914, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0016778760007582604, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.009962321212515234, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.015294800698757171, + "weighted_rkd_loss": 0.004981160606257617 + }, + { + "epoch": 8.205128205128204, + "flow_matching_loss": 0.016772223822772502, + "grad_norm": 0.161229208111763, + "learning_rate": 1.323506904413485e-07, + "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": 0.003599933069199324, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0016003339667804539, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.008800200140103697, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.016772223822772502, + "weighted_rkd_loss": 0.0044001000700518485 + }, + { + "epoch": 8.222222222222221, + "flow_matching_loss": 0.014597597904503346, + "grad_norm": 0.15218256413936615, + "learning_rate": 1.195261433903594e-07, + "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": 0.004443337768316269, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0016796654323115944, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.010566340759396553, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.014597597904503346, + "weighted_rkd_loss": 0.0052831703796982765 + }, + { + "epoch": 8.239316239316238, + "flow_matching_loss": 0.013690507505089045, + "grad_norm": 0.1815441995859146, + "learning_rate": 1.0735251647190969e-07, + "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": 0.004198707919567823, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0016398903797380626, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.010037306183949112, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.013690507505089045, + "weighted_rkd_loss": 0.005018653091974556 + }, + { + "epoch": 8.256410256410255, + "flow_matching_loss": 0.01400683168321848, + "grad_norm": 0.18531562387943268, + "learning_rate": 9.583034219987408e-08, + "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": 0.0039456434780731795, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0015729700447991491, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.009464256884530186, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.01400683168321848, + "weighted_rkd_loss": 0.004732128442265093 + }, + { + "epoch": 8.273504273504274, + "flow_matching_loss": 0.013479845225811004, + "grad_norm": 0.12994004786014557, + "learning_rate": 8.496012459148717e-08, + "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": 0.00363312978297472, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0013098492461722344, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.008576108841225505, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.013479845225811004, + "weighted_rkd_loss": 0.0042880544206127524 + }, + { + "epoch": 8.290598290598291, + "flow_matching_loss": 0.01754144886508584, + "grad_norm": 0.14384311437606812, + "learning_rate": 7.474233914527618e-08, + "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": 0.003767120442353189, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0014266954676713794, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.008960936358198524, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.01754144886508584, + "weighted_rkd_loss": 0.004480468179099262 + }, + { + "epoch": 8.307692307692308, + "flow_matching_loss": 0.017515084147453307, + "grad_norm": 0.14004142582416534, + "learning_rate": 6.517743282028254e-08, + "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": 0.004615483921952546, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0017025555716827512, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.010933523252606391, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.017515084147453307, + "weighted_rkd_loss": 0.005466761626303196 + }, + { + "epoch": 8.324786324786325, + "flow_matching_loss": 0.01451965318992734, + "grad_norm": 0.18907830119132996, + "learning_rate": 5.6265824016492604e-08, + "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": 0.004115891247056425, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0019405572209507227, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.010172339901328088, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.01451965318992734, + "weighted_rkd_loss": 0.005086169950664044 + }, + { + "epoch": 8.341880341880342, + "flow_matching_loss": 0.019001572486013175, + "grad_norm": 0.20190921425819397, + "learning_rate": 4.800790255654719e-08, + "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": 0.004398517566733063, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0019266181974671782, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.010723653389140964, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.019001572486013175, + "weighted_rkd_loss": 0.005361826694570482 + }, + { + "epoch": 8.35897435897436, + "flow_matching_loss": 0.015850430447608234, + "grad_norm": 0.1569550633430481, + "learning_rate": 4.040402966867862e-08, + "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": 0.005174251366406679, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.001973703713156283, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.012322206469252705, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.015850430447608234, + "weighted_rkd_loss": 0.0061611032346263524 + }, + { + "epoch": 8.376068376068377, + "flow_matching_loss": 0.01483481116592884, + "grad_norm": 0.15155529975891113, + "learning_rate": 3.3454537970925014e-08, + "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": 0.004053276544436812, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0016709259594790637, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.009777478920295835, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.01483481116592884, + "weighted_rkd_loss": 0.004888739460147918 + }, + { + "epoch": 8.393162393162394, + "flow_matching_loss": 0.018385430797934532, + "grad_norm": 0.17632077634334564, + "learning_rate": 2.715973145656192e-08, + "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": 0.003842594940215349, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0014750556205399335, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.009160245349630714, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.018385430797934532, + "weighted_rkd_loss": 0.004580122674815357 + }, + { + "epoch": 8.41025641025641, + "flow_matching_loss": 0.014689198601990939, + "grad_norm": 0.1562715321779251, + "learning_rate": 2.151988548082129e-08, + "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": 0.004287013225257397, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0018384936323855073, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.010412520170211792, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.014689198601990939, + "weighted_rkd_loss": 0.005206260085105896 + }, + { + "epoch": 8.427350427350428, + "flow_matching_loss": 0.01542881028726697, + "grad_norm": 0.17578746378421783, + "learning_rate": 1.6535246748837772e-08, + "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": 0.003949931729584933, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.001388546999078244, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.009288410376757383, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.01542881028726697, + "weighted_rkd_loss": 0.004644205188378691 + }, + { + "epoch": 8.444444444444445, + "flow_matching_loss": 0.01600944958627224, + "grad_norm": 0.1673750877380371, + "learning_rate": 1.2206033304855702e-08, + "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": 0.0038427982944995165, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0016363873146474361, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.00932198385708034, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.01600944958627224, + "weighted_rkd_loss": 0.00466099192854017 + }, + { + "epoch": 8.461538461538462, + "flow_matching_loss": 0.016356547176837922, + "grad_norm": 0.15482042729854584, + "learning_rate": 8.532434522695031e-09, + "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": 0.004332413501106203, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.001804306684061885, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.010469133732840419, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.016356547176837922, + "weighted_rkd_loss": 0.005234566866420209 + }, + { + "epoch": 8.478632478632479, + "flow_matching_loss": 0.01732535157352686, + "grad_norm": 0.16185925900936127, + "learning_rate": 5.5146110974679805e-09, + "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": 0.004459207039326429, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.001728048489894718, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.010646462440490723, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.01732535157352686, + "weighted_rkd_loss": 0.005323231220245361 + }, + { + "epoch": 8.495726495726496, + "flow_matching_loss": 0.015061072912067175, + "grad_norm": 0.23401165008544922, + "learning_rate": 3.1526950385496467e-09, + "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": 0.00431763231754303, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0017549702781252563, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01039023487828672, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.015061072912067175, + "weighted_rkd_loss": 0.00519511743914336 + }, + { + "epoch": 8.512820512820513, + "flow_matching_loss": 0.013777593802660703, + "grad_norm": 0.2139202058315277, + "learning_rate": 1.4467896637959666e-09, + "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": 0.00465741497464478, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0016942919231951238, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.011009121919050813, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.013777593802660703, + "weighted_rkd_loss": 0.0055045609595254065 + }, + { + "epoch": 8.52991452991453, + "flow_matching_loss": 0.015110108815133572, + "grad_norm": 0.13214650750160217, + "learning_rate": 3.9696959503898733e-10, + "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": 0.003758502355776727, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.00138128298567608, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.00889828773215413, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.015110108815133572, + "weighted_rkd_loss": 0.004449143866077065 + }, + { + "epoch": 8.547008547008547, + "flow_matching_loss": 0.014968675933778286, + "grad_norm": 0.17623348534107208, + "learning_rate": 3.280754806156416e-12, + "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": 0.004165530833415687, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0017792441823985428, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01011030594818294, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.014968675933778286, + "weighted_rkd_loss": 0.00505515297409147 + }, + { + "epoch": 8.547008547008547, + "step": 5000, + "total_flos": 0.0, + "train_loss": 0.042555030176043514, + "train_runtime": 8961.6957, + "train_samples_per_second": 35.708, + "train_steps_per_second": 0.558 + } + ], + "logging_steps": 10, + "max_steps": 5000, + "num_input_tokens_seen": 0, + "num_train_epochs": 9, + "save_steps": 5001, + "stateful_callbacks": { + "TrainerControl": { + "args": { + "should_epoch_stop": false, + "should_evaluate": false, + "should_log": false, + "should_save": true, + "should_training_stop": true + }, + "attributes": {} + } + }, + "total_flos": 0.0, + "train_batch_size": 64, + "trial_name": null, + "trial_params": null +} diff --git a/FewShot/39tasks/5shot/FromRKD-Timewarp_FullactionHead/seed0/default/resolved_config.yaml b/FewShot/39tasks/5shot/FromRKD-Timewarp_FullactionHead/seed0/default/resolved_config.yaml new file mode 100644 index 0000000000000000000000000000000000000000..ba0bd4def2025cbf94fbaa420181442647b9988d --- /dev/null +++ b/FewShot/39tasks/5shot/FromRKD-Timewarp_FullactionHead/seed0/default/resolved_config.yaml @@ -0,0 +1,38 @@ +experiment_name: fewshot_39tasks_5shot_from_rkd_timewarp_fullactionhead_seed0 +policy_type: groot_RKDv3 +base_model_path: ${HOME}/groot_robocasa/robocasa_v2/TimewarpVAE/RKDv3_DA_APHead_from_Atomic26Baseline60k/default/phase3 +output_root: ${HOME}/groot_robocasa/robocasa_v2/FewShot/39tasks/5shot/FromRKD-Timewarp_FullactionHead/seed0 +dataset: + dataset_soup: fewshot_heldout_atomic39_human_5shot +training: + num_gpus: 1 + batch_size: 64 + seed: 42 + filter_key_seed: 0 + manifest_path: ${HOME}/clvla/benchmarks/robocasa_v2/my_scripts/Fewshot/manifests/fewshot_heldout_atomic39/shot5_seed0.json + save_total_limit: 4 +model: + rkd_enabled: true + rkd_fm_loss_weight: 1.0 + rkd_loss_weight: 0.5 + rkd_loss_weight_schedule: null +phases: +- name: train_5shot_seed0_fullactionhead_5k + max_steps: 5000 + save_steps: 0 + save_step_list: + - 1000 + - 3000 + - 5000 + save_final_model: false + 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: 0.5 + rkd_loss_weight_schedule: null +resolved_sweep: {} diff --git a/FewShot/39tasks/5shot/FromRKD-Timewarp_FullactionHead/seed0/default/seed0.yaml b/FewShot/39tasks/5shot/FromRKD-Timewarp_FullactionHead/seed0/default/seed0.yaml new file mode 100644 index 0000000000000000000000000000000000000000..fa7ee6b14f883adb3c74459cb8e20bbce6111d6f --- /dev/null +++ b/FewShot/39tasks/5shot/FromRKD-Timewarp_FullactionHead/seed0/default/seed0.yaml @@ -0,0 +1,37 @@ +experiment_name: fewshot_39tasks_5shot_from_rkd_timewarp_fullactionhead_seed0 +policy_type: groot_RKDv3 +base_model_path: ${HOME}/groot_robocasa/robocasa_v2/TimewarpVAE/RKDv3_DA_APHead_from_Atomic26Baseline60k/default/phase3 +output_root: ${HOME}/groot_robocasa/robocasa_v2/FewShot/39tasks/5shot/FromRKD-Timewarp_FullactionHead/seed0 +dataset: + dataset_soup: fewshot_heldout_atomic39_human_5shot +training: + num_gpus: 1 + batch_size: 64 + seed: 42 + filter_key_seed: 0 + manifest_path: ${HOME}/clvla/benchmarks/robocasa_v2/my_scripts/Fewshot/manifests/fewshot_heldout_atomic39/shot5_seed0.json + save_total_limit: 4 +model: + rkd_enabled: true + rkd_fm_loss_weight: 1.0 + rkd_loss_weight: 0.5 + rkd_loss_weight_schedule: null +phases: +- name: train_5shot_seed0_fullactionhead_5k + max_steps: 5000 + save_steps: 0 + save_step_list: + - 1000 + - 3000 + - 5000 + save_final_model: false + 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: 0.5 + rkd_loss_weight_schedule: null diff --git a/FewShot/39tasks/5shot/FromRKD-Timewarp_FullactionHead/seed0/default/train_5shot_seed0_fullactionhead_5k/checkpoint-1000/config.json b/FewShot/39tasks/5shot/FromRKD-Timewarp_FullactionHead/seed0/default/train_5shot_seed0_fullactionhead_5k/checkpoint-1000/config.json new file mode 100644 index 0000000000000000000000000000000000000000..4740d0455bef31c469ee5dab6d4a47d64df61665 --- /dev/null +++ b/FewShot/39tasks/5shot/FromRKD-Timewarp_FullactionHead/seed0/default/train_5shot_seed0_fullactionhead_5k/checkpoint-1000/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": 0.5, + "rkd_loss_weight_end": 0.0, + "rkd_loss_weight_schedule": null, + "rkd_loss_weight_start": 0.1, + "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/FewShot/39tasks/5shot/FromRKD-Timewarp_FullactionHead/seed0/default/train_5shot_seed0_fullactionhead_5k/checkpoint-1000/experiment_cfg/metadata.json b/FewShot/39tasks/5shot/FromRKD-Timewarp_FullactionHead/seed0/default/train_5shot_seed0_fullactionhead_5k/checkpoint-1000/experiment_cfg/metadata.json new file mode 100644 index 0000000000000000000000000000000000000000..6f73f8489294c58acc82d1351a17e5e3ff316e18 --- /dev/null +++ b/FewShot/39tasks/5shot/FromRKD-Timewarp_FullactionHead/seed0/default/train_5shot_seed0_fullactionhead_5k/checkpoint-1000/experiment_cfg/metadata.json @@ -0,0 +1,431 @@ +{ + "new_embodiment": { + "statistics": { + "state": { + "base_position": { + "max": [ + 7.778976917266846, + 0.33403685688972473, + 0.7223060727119446 + ], + "min": [ + -4.821934223175049, + -6.785110950469971, + 0.6994727253913879 + ], + "mean": [ + 2.5886094736594543, + -2.1559664577316586, + 0.7003580854548973 + ], + "std": [ + 1.6717149343420803, + 1.3743695399555862, + 0.000940069542359506 + ], + "q01": [ + -0.7517815060249613, + -5.497849509089705, + 0.6999776914424374 + ], + "q99": [ + 6.060945690807028, + -0.5331499932493624, + 0.7028864260472257 + ] + }, + "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, + 0.24488939171811902, + 0.5968549204077488 + ], + "std": [ + 0.0, + 0.0, + 0.6659970734507764, + 0.37443298392262264 + ], + "q01": [ + 0.0, + 0.0, + -0.9999999999999999, + 4.149924358336069e-06 + ], + "q99": [ + 0.0, + 0.0, + 0.9999999999999999, + 0.9999999999999999 + ] + }, + "end_effector_position_relative": { + "max": [ + 0.8848381638526917, + 0.8569459915161133, + 1.053884744644165 + ], + "min": [ + -0.5828549861907959, + -0.8486236333847046, + -0.3446122705936432 + ], + "mean": [ + 0.25087486209693133, + -0.005738920462036393, + 0.43643948866361787 + ], + "std": [ + 0.15741544532518012, + 0.2547603106793105, + 0.2050434492495558 + ], + "q01": [ + -0.013990057288082561, + -0.4824599092623036, + 0.18825220353500022 + ], + "q99": [ + 0.5411989642841469, + 0.4754923657245658, + 0.6912323747882034 + ] + }, + "end_effector_rotation_relative": { + "max": [ + 0.9999949932098389, + 0.9997354745864868, + 0.9900747537612915, + 0.9735639095306396 + ], + "min": [ + -0.9999914169311523, + -0.9999884366989136, + -0.9975369572639465, + 6.754255821306288e-08 + ], + "mean": [ + -0.22290331318740714, + 0.04579037037503981, + -0.05804094906527823, + 0.12347219024917105 + ], + "std": [ + 0.8519267931761344, + 0.2949394732270141, + 0.30026078832556913, + 0.16333086331829588 + ], + "q01": [ + -0.9957959450418518, + -0.5678869493126506, + -0.5318135475761875, + 0.0014160465686576505 + ], + "q99": [ + 0.9900692296898317, + 0.6357358195673684, + 0.5009847868015971, + 0.4287044482210845 + ] + }, + "gripper_qpos": { + "max": [ + 0.05289580300450325, + 0.00937709677964449 + ], + "min": [ + -0.011669117026031017, + -0.06307756155729294 + ], + "mean": [ + 0.03203268930378034, + -0.032331639271968075 + ], + "std": [ + 0.011607049173153697, + 0.011306286745872529 + ], + "q01": [ + 0.006429109393656672, + -0.04056586943958638 + ], + "q99": [ + 0.040377860142451545, + -0.006593889698673258 + ] + } + }, + "action": { + "base_motion": { + "max": [ + 1.0, + 1.0, + 1.0, + 0.0 + ], + "min": [ + -1.0, + -1.0, + -1.0, + 0.0 + ], + "mean": [ + 0.006050336852760097, + 0.002728916466628117, + -0.0017236811393598838, + 0.0 + ], + "std": [ + 0.11032167711449951, + 0.09824140301615687, + 0.07825208047515658, + 0.0 + ], + "q01": [ + -0.154721262094005, + -0.18955302988697728, + -0.14454065867754942, + 0.0 + ], + "q99": [ + 0.24492955247744197, + 0.2496797822190429, + 0.11966742325336034, + 0.0 + ] + }, + "control_mode": { + "max": [ + 1.0 + ], + "min": [ + -1.0 + ], + "mean": [ + -0.9294448715752568 + ], + "std": [ + 0.36897903503358326 + ], + "q01": [ + -1.0 + ], + "q99": [ + -0.21627562513311385 + ] + }, + "end_effector_position": { + "max": [ + 1.0, + 1.0, + 1.0 + ], + "min": [ + -1.0, + -1.0, + -1.0 + ], + "mean": [ + -0.00826780165926883, + 0.003161997359379232, + -0.08157453651841007 + ], + "std": [ + 0.4545430944464769, + 0.41040151434986594, + 0.4102644251536544 + ], + "q01": [ + -0.9653185255999107, + -0.9255635199318754, + -0.8208437313341718 + ], + "q99": [ + 0.7987781084592441, + 0.9260844232330088, + 0.8357421812622201 + ] + }, + "end_effector_rotation": { + "max": [ + 1.0, + 1.0, + 1.0 + ], + "min": [ + -1.0, + -1.0, + -1.0 + ], + "mean": [ + 0.007367024641295137, + -0.02060756343728361, + -0.001884833075998489 + ], + "std": [ + 0.1116768956941076, + 0.13249042272165212, + 0.12338908364725322 + ], + "q01": [ + -0.26894658663246845, + -0.39865607740339326, + -0.34393735191302655 + ], + "q99": [ + 0.3191674780171587, + 0.31133630397431267, + 0.33968476423051097 + ] + }, + "gripper_close": { + "max": [ + 1.0 + ], + "min": [ + -1.0 + ], + "mean": [ + -0.3098994894722959 + ], + "std": [ + 0.9507718287498876 + ], + "q01": [ + -1.0 + ], + "q99": [ + 0.7545620899936146 + ] + } + } + }, + "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/FewShot/39tasks/5shot/FromRKD-Timewarp_FullactionHead/seed0/default/train_5shot_seed0_fullactionhead_5k/checkpoint-1000/model.safetensors.index.json b/FewShot/39tasks/5shot/FromRKD-Timewarp_FullactionHead/seed0/default/train_5shot_seed0_fullactionhead_5k/checkpoint-1000/model.safetensors.index.json new file mode 100644 index 0000000000000000000000000000000000000000..b20baed71f4d16a411223b70d45d934c6cbbfe97 --- /dev/null +++ b/FewShot/39tasks/5shot/FromRKD-Timewarp_FullactionHead/seed0/default/train_5shot_seed0_fullactionhead_5k/checkpoint-1000/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/FewShot/39tasks/5shot/FromRKD-Timewarp_FullactionHead/seed0/default/train_5shot_seed0_fullactionhead_5k/checkpoint-1000/trainer_state.json b/FewShot/39tasks/5shot/FromRKD-Timewarp_FullactionHead/seed0/default/train_5shot_seed0_fullactionhead_5k/checkpoint-1000/trainer_state.json new file mode 100644 index 0000000000000000000000000000000000000000..6cb92eec0bb0aa481ad37274e954512d4fc5c315 --- /dev/null +++ b/FewShot/39tasks/5shot/FromRKD-Timewarp_FullactionHead/seed0/default/train_5shot_seed0_fullactionhead_5k/checkpoint-1000/trainer_state.json @@ -0,0 +1,4634 @@ +{ + "best_global_step": null, + "best_metric": null, + "best_model_checkpoint": null, + "epoch": 1.6778523489932886, + "eval_steps": 500, + "global_step": 1000, + "is_hyper_param_search": false, + "is_local_process_zero": true, + "is_world_process_zero": true, + "log_history": [ + { + "epoch": 0.016778523489932886, + "flow_matching_loss": 0.189423368871212, + "grad_norm": 1.1502586603164673, + "learning_rate": 1.08e-06, + "loss": 0.38, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.13627509251236916, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.10864389836788177, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.38119408786296843, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.189423368871212, + "weighted_rkd_loss": 0.19059704393148422 + }, + { + "epoch": 0.03355704697986577, + "flow_matching_loss": 0.21835082918405532, + "grad_norm": 1.3782790899276733, + "learning_rate": 2.28e-06, + "loss": 0.4112, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.13762855604290963, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.11038857251405716, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.3856456845998764, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.21835082918405532, + "weighted_rkd_loss": 0.1928228422999382 + }, + { + "epoch": 0.050335570469798654, + "flow_matching_loss": 0.1594226397573948, + "grad_norm": 1.0711982250213623, + "learning_rate": 3.48e-06, + "loss": 0.3382, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.1283317506313324, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.10086342692375183, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.3575269281864166, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.1594226397573948, + "weighted_rkd_loss": 0.1787634640932083 + }, + { + "epoch": 0.06711409395973154, + "flow_matching_loss": 0.15615404024720192, + "grad_norm": 0.7945505380630493, + "learning_rate": 4.68e-06, + "loss": 0.3226, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.12152766585350036, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0897677019238472, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.33282303512096406, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.15615404024720192, + "weighted_rkd_loss": 0.16641151756048203 + }, + { + "epoch": 0.08389261744966443, + "flow_matching_loss": 0.13071623146533967, + "grad_norm": 0.8744078278541565, + "learning_rate": 5.8800000000000005e-06, + "loss": 0.2952, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.11974534913897514, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.08949827328324318, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.32898896336555483, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.13071623146533967, + "weighted_rkd_loss": 0.16449448168277742 + }, + { + "epoch": 0.10067114093959731, + "flow_matching_loss": 0.11403126865625382, + "grad_norm": 0.651993453502655, + "learning_rate": 7.08e-06, + "loss": 0.2721, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.11685370355844497, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.08247915208339691, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.3161865621805191, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.11403126865625382, + "weighted_rkd_loss": 0.15809328109025955 + }, + { + "epoch": 0.1174496644295302, + "flow_matching_loss": 0.0997005932033062, + "grad_norm": 0.6273358464241028, + "learning_rate": 8.28e-06, + "loss": 0.2453, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.10773090794682502, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.07566531300544739, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.29112712740898133, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.0997005932033062, + "weighted_rkd_loss": 0.14556356370449067 + }, + { + "epoch": 0.1342281879194631, + "flow_matching_loss": 0.08913109861314297, + "grad_norm": 0.6481125950813293, + "learning_rate": 9.48e-06, + "loss": 0.2235, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.10180731341242791, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.06521718949079514, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.2688318148255348, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.08913109861314297, + "weighted_rkd_loss": 0.1344159074127674 + }, + { + "epoch": 0.15100671140939598, + "flow_matching_loss": 0.07915030717849732, + "grad_norm": 0.6114252209663391, + "learning_rate": 1.068e-05, + "loss": 0.2091, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.09674249440431595, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.06643815003335476, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.2599231392145157, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.07915030717849732, + "weighted_rkd_loss": 0.12996156960725785 + }, + { + "epoch": 0.16778523489932887, + "flow_matching_loss": 0.07286151051521302, + "grad_norm": 0.5393626093864441, + "learning_rate": 1.1880000000000001e-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": 0.08473351448774338, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.054971330985426906, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.2244383603334427, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.07286151051521302, + "weighted_rkd_loss": 0.11221918016672135 + }, + { + "epoch": 0.18456375838926176, + "flow_matching_loss": 0.07065574415028095, + "grad_norm": 0.6259891390800476, + "learning_rate": 1.308e-05, + "loss": 0.1874, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.08815310075879097, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.05715718045830727, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.23346338123083116, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.07065574415028095, + "weighted_rkd_loss": 0.11673169061541558 + }, + { + "epoch": 0.20134228187919462, + "flow_matching_loss": 0.0647046722471714, + "grad_norm": 0.7180153131484985, + "learning_rate": 1.428e-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": 0.07692675888538361, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.05016751065850258, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.20402102917432785, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.0647046722471714, + "weighted_rkd_loss": 0.10201051458716393 + }, + { + "epoch": 0.2181208053691275, + "flow_matching_loss": 0.06340371444821358, + "grad_norm": 0.5939756035804749, + "learning_rate": 1.548e-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": 0.07837383523583412, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.04749012887477875, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.20423779934644698, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.06340371444821358, + "weighted_rkd_loss": 0.10211889967322349 + }, + { + "epoch": 0.2348993288590604, + "flow_matching_loss": 0.06330732218921184, + "grad_norm": 0.5114864110946655, + "learning_rate": 1.6680000000000003e-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": 0.07281922809779644, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.04322611875832081, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.1888645753264427, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.06330732218921184, + "weighted_rkd_loss": 0.09443228766322136 + }, + { + "epoch": 0.2516778523489933, + "flow_matching_loss": 0.05951428152620793, + "grad_norm": 0.6290663480758667, + "learning_rate": 1.7879999999999998e-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": 0.06923839673399926, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.041895052045583726, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.18037184625864028, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.05951428152620793, + "weighted_rkd_loss": 0.09018592312932014 + }, + { + "epoch": 0.2684563758389262, + "flow_matching_loss": 0.058728757128119466, + "grad_norm": 0.5164991021156311, + "learning_rate": 1.908e-05, + "loss": 0.1374, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.060720610246062276, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0359353007748723, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.15737652033567429, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.058728757128119466, + "weighted_rkd_loss": 0.07868826016783714 + }, + { + "epoch": 0.28523489932885904, + "flow_matching_loss": 0.05690980330109596, + "grad_norm": 0.42798617482185364, + "learning_rate": 2.0280000000000002e-05, + "loss": 0.1344, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.05981909520924091, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.03533888217061758, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.1549770712852478, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.05690980330109596, + "weighted_rkd_loss": 0.0774885356426239 + }, + { + "epoch": 0.30201342281879195, + "flow_matching_loss": 0.05503378733992577, + "grad_norm": 0.4299779534339905, + "learning_rate": 2.148e-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": 0.05938076414167881, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.03561958894133568, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.15438111796975135, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.05503378733992577, + "weighted_rkd_loss": 0.07719055898487567 + }, + { + "epoch": 0.3187919463087248, + "flow_matching_loss": 0.054644962027668956, + "grad_norm": 0.4563824534416199, + "learning_rate": 2.268e-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": 0.05716169141232967, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.03335041459649801, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.147673799097538, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.054644962027668956, + "weighted_rkd_loss": 0.073836899548769 + }, + { + "epoch": 0.33557046979865773, + "flow_matching_loss": 0.056772416457533836, + "grad_norm": 0.4835961163043976, + "learning_rate": 2.3880000000000002e-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": 0.054468106105923654, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.03251056261360645, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.14144677594304084, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.056772416457533836, + "weighted_rkd_loss": 0.07072338797152042 + }, + { + "epoch": 0.3523489932885906, + "flow_matching_loss": 0.0495440598577261, + "grad_norm": 0.5092470049858093, + "learning_rate": 2.508e-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": 0.04748903773725033, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.026358308084309102, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.12133638188242912, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.0495440598577261, + "weighted_rkd_loss": 0.06066819094121456 + }, + { + "epoch": 0.3691275167785235, + "flow_matching_loss": 0.0511940747499466, + "grad_norm": 0.4401130974292755, + "learning_rate": 2.628e-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": 0.0482368066906929, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0273994205519557, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.1238730326294899, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.0511940747499466, + "weighted_rkd_loss": 0.06193651631474495 + }, + { + "epoch": 0.3859060402684564, + "flow_matching_loss": 0.050138309597969055, + "grad_norm": 0.42907798290252686, + "learning_rate": 2.748e-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": 0.0430266585201025, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.025129231996834277, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.11118254959583282, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.050138309597969055, + "weighted_rkd_loss": 0.05559127479791641 + }, + { + "epoch": 0.40268456375838924, + "flow_matching_loss": 0.046704120188951495, + "grad_norm": 0.4718972146511078, + "learning_rate": 2.868e-05, + "loss": 0.1016, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.042805034294724466, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.024135740101337434, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.10974581018090249, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.046704120188951495, + "weighted_rkd_loss": 0.05487290509045124 + }, + { + "epoch": 0.41946308724832215, + "flow_matching_loss": 0.044243189692497256, + "grad_norm": 0.3790624737739563, + "learning_rate": 2.9880000000000002e-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": 0.041216658987104894, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.021118617989122866, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.10355193763971329, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.044243189692497256, + "weighted_rkd_loss": 0.051775968819856646 + }, + { + "epoch": 0.436241610738255, + "flow_matching_loss": 0.04542431682348251, + "grad_norm": 0.512637197971344, + "learning_rate": 2.9999734259635676e-05, + "loss": 0.0959, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.03973671272397041, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.021489275991916655, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.10096270069479943, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.04542431682348251, + "weighted_rkd_loss": 0.050481350347399714 + }, + { + "epoch": 0.45302013422818793, + "flow_matching_loss": 0.05049298815429211, + "grad_norm": 0.4598679840564728, + "learning_rate": 2.9998815663057245e-05, + "loss": 0.102, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.040620113164186476, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.021747449319809674, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.10298767462372779, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.05049298815429211, + "weighted_rkd_loss": 0.051493837311863896 + }, + { + "epoch": 0.4697986577181208, + "flow_matching_loss": 0.046248992905020714, + "grad_norm": 0.4032992422580719, + "learning_rate": 2.999724096969228e-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": 0.038419022038578984, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.021234015561640264, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.09807205945253372, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.046248992905020714, + "weighted_rkd_loss": 0.04903602972626686 + }, + { + "epoch": 0.4865771812080537, + "flow_matching_loss": 0.04581176079809666, + "grad_norm": 0.378512978553772, + "learning_rate": 2.9995010248422984e-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": 0.036053233221173285, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.019097956269979476, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.09120442345738411, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.04581176079809666, + "weighted_rkd_loss": 0.045602211728692055 + }, + { + "epoch": 0.5033557046979866, + "flow_matching_loss": 0.040904562920331955, + "grad_norm": 0.3660767078399658, + "learning_rate": 2.999212359682832e-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": 0.030510597676038743, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.015882878191769124, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.07690407261252404, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.040904562920331955, + "weighted_rkd_loss": 0.03845203630626202 + }, + { + "epoch": 0.5201342281879194, + "flow_matching_loss": 0.03941964264959097, + "grad_norm": 0.3793014883995056, + "learning_rate": 2.9988581141179787e-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": 0.032410606555640695, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.01629540789872408, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.08111662119626999, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.03941964264959097, + "weighted_rkd_loss": 0.04055831059813499 + }, + { + "epoch": 0.5369127516778524, + "flow_matching_loss": 0.04001129996031523, + "grad_norm": 0.6170795559883118, + "learning_rate": 2.998438303643587e-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": 0.033591178432106975, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.018490686919540165, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.08567304462194443, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.04001129996031523, + "weighted_rkd_loss": 0.042836522310972215 + }, + { + "epoch": 0.5536912751677853, + "flow_matching_loss": 0.040321177989244464, + "grad_norm": 0.3570716381072998, + "learning_rate": 2.997952946623527e-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": 0.031730389967560765, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.016793423146009446, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.08025420233607292, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.040321177989244464, + "weighted_rkd_loss": 0.04012710116803646 + }, + { + "epoch": 0.5704697986577181, + "flow_matching_loss": 0.04315701965242624, + "grad_norm": 0.379740446805954, + "learning_rate": 2.9974020642888875e-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": 0.03032686710357666, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.01704806601628661, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.07770180031657219, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.04315701965242624, + "weighted_rkd_loss": 0.03885090015828609 + }, + { + "epoch": 0.587248322147651, + "flow_matching_loss": 0.040681138448417184, + "grad_norm": 0.33822014927864075, + "learning_rate": 2.9967856807370458e-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": 0.027930348366498946, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0157440728507936, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.0716047689318657, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.040681138448417184, + "weighted_rkd_loss": 0.03580238446593285 + }, + { + "epoch": 0.6040268456375839, + "flow_matching_loss": 0.042508364841341975, + "grad_norm": 0.40825992822647095, + "learning_rate": 2.996103822930615e-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": 0.031780967675149444, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.016943857539445162, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.08050579428672791, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.042508364841341975, + "weighted_rkd_loss": 0.040252897143363955 + }, + { + "epoch": 0.6208053691275168, + "flow_matching_loss": 0.037045056186616424, + "grad_norm": 0.3486769199371338, + "learning_rate": 2.9953565206962653e-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": 0.028993065282702447, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.014874111022800207, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.07286024205386639, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.037045056186616424, + "weighted_rkd_loss": 0.036430121026933195 + }, + { + "epoch": 0.6375838926174496, + "flow_matching_loss": 0.038407695665955546, + "grad_norm": 0.4263972342014313, + "learning_rate": 2.9945438067234172e-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": 0.029730910249054433, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.015382550936192275, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.07484437264502049, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.038407695665955546, + "weighted_rkd_loss": 0.037422186322510244 + }, + { + "epoch": 0.6543624161073825, + "flow_matching_loss": 0.038366636447608474, + "grad_norm": 0.3533881604671478, + "learning_rate": 2.9936657165628124e-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": 0.027479714527726174, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.014265698939561844, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.0692251279950142, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.038366636447608474, + "weighted_rkd_loss": 0.0346125639975071 + }, + { + "epoch": 0.6711409395973155, + "flow_matching_loss": 0.035504350997507574, + "grad_norm": 0.3166177272796631, + "learning_rate": 2.992722288624959e-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": 0.025187848880887032, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.013730826880782842, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.06410652436316014, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.035504350997507574, + "weighted_rkd_loss": 0.03205326218158007 + }, + { + "epoch": 0.6879194630872483, + "flow_matching_loss": 0.037043062038719654, + "grad_norm": 0.3390388488769531, + "learning_rate": 2.9917135641784515e-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": 0.0239717124029994, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.012146859616041183, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.060090285167098044, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.037043062038719654, + "weighted_rkd_loss": 0.030045142583549022 + }, + { + "epoch": 0.7046979865771812, + "flow_matching_loss": 0.03493194635957479, + "grad_norm": 0.3506368398666382, + "learning_rate": 2.9906395873481655e-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": 0.025168376229703426, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.012817893270403146, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.06315464563667775, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.03493194635957479, + "weighted_rkd_loss": 0.031577322818338874 + }, + { + "epoch": 0.7214765100671141, + "flow_matching_loss": 0.036136615090072155, + "grad_norm": 0.39711323380470276, + "learning_rate": 2.989500405113326e-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": 0.025528428331017494, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.012982557248324155, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.06403941474854946, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.036136615090072155, + "weighted_rkd_loss": 0.03201970737427473 + }, + { + "epoch": 0.738255033557047, + "flow_matching_loss": 0.0340481685474515, + "grad_norm": 0.3472777009010315, + "learning_rate": 2.9882960673054547e-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": 0.024420114979147912, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.012645702250301838, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.061485931649804114, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.0340481685474515, + "weighted_rkd_loss": 0.030742965824902057 + }, + { + "epoch": 0.7550335570469798, + "flow_matching_loss": 0.035830936022102834, + "grad_norm": 0.34199073910713196, + "learning_rate": 2.9870266266061886e-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": 0.02066250964999199, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.01092346548102796, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.05224848538637161, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.035830936022102834, + "weighted_rkd_loss": 0.026124242693185806 + }, + { + "epoch": 0.7718120805369127, + "flow_matching_loss": 0.03825925439596176, + "grad_norm": 0.3056429326534271, + "learning_rate": 2.9856921385449766e-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": 0.026111494936048984, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.01336317528039217, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.06558616384863854, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.03825925439596176, + "weighted_rkd_loss": 0.03279308192431927 + }, + { + "epoch": 0.7885906040268457, + "flow_matching_loss": 0.03638801425695419, + "grad_norm": 0.3128032982349396, + "learning_rate": 2.9842926614966502e-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": 0.02268179152160883, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.011993827018886804, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.05735741071403026, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.03638801425695419, + "weighted_rkd_loss": 0.02867870535701513 + }, + { + "epoch": 0.8053691275167785, + "flow_matching_loss": 0.03183348029851914, + "grad_norm": 0.2547442317008972, + "learning_rate": 2.982828256678868e-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": 0.022423543594777585, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.011085591791197658, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.05593267902731895, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.03183348029851914, + "weighted_rkd_loss": 0.027966339513659476 + }, + { + "epoch": 0.8221476510067114, + "flow_matching_loss": 0.03268085308372974, + "grad_norm": 0.2893790006637573, + "learning_rate": 2.9812989881494412e-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": 0.02247578836977482, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.011517873685806989, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.05646945089101792, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.03268085308372974, + "weighted_rkd_loss": 0.02823472544550896 + }, + { + "epoch": 0.8389261744966443, + "flow_matching_loss": 0.038189934380352494, + "grad_norm": 0.4075469672679901, + "learning_rate": 2.9797049228035296e-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": 0.02296894444152713, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.012043415708467365, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.05798130407929421, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.038189934380352494, + "weighted_rkd_loss": 0.028990652039647104 + }, + { + "epoch": 0.8557046979865772, + "flow_matching_loss": 0.03394890353083611, + "grad_norm": 0.36510178446769714, + "learning_rate": 2.978046130370715e-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": 0.02077607475221157, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.01056032245978713, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.0521124716848135, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.03394890353083611, + "weighted_rkd_loss": 0.02605623584240675 + }, + { + "epoch": 0.87248322147651, + "flow_matching_loss": 0.03357741795480251, + "grad_norm": 0.2992410361766815, + "learning_rate": 2.9763226834119535e-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": 0.020533321145921945, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.010199433751404285, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.05126607567071915, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.03357741795480251, + "weighted_rkd_loss": 0.025633037835359574 + }, + { + "epoch": 0.889261744966443, + "flow_matching_loss": 0.03181673102080822, + "grad_norm": 0.30052074790000916, + "learning_rate": 2.9745346573163963e-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": 0.018613563012331726, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.009025087859481574, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.04625221379101276, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.03181673102080822, + "weighted_rkd_loss": 0.02312610689550638 + }, + { + "epoch": 0.9060402684563759, + "flow_matching_loss": 0.031556636840105054, + "grad_norm": 0.28555840253829956, + "learning_rate": 2.9726821302980987e-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": 0.018975557200610636, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.010126672731712461, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.048077787086367604, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.031556636840105054, + "weighted_rkd_loss": 0.024038893543183802 + }, + { + "epoch": 0.9228187919463087, + "flow_matching_loss": 0.031900867260992526, + "grad_norm": 0.3493785560131073, + "learning_rate": 2.9707651833925935e-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": 0.01944209486246109, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.00898348679766059, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.047867676615715025, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.031900867260992526, + "weighted_rkd_loss": 0.023933838307857513 + }, + { + "epoch": 0.9395973154362416, + "flow_matching_loss": 0.033641642332077025, + "grad_norm": 0.3126336932182312, + "learning_rate": 2.968783900453348e-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": 0.018097881507128478, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.009972660942003131, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.04616842418909073, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.033641642332077025, + "weighted_rkd_loss": 0.023084212094545364 + }, + { + "epoch": 0.9563758389261745, + "flow_matching_loss": 0.03296861108392477, + "grad_norm": 0.3514993488788605, + "learning_rate": 2.9667383681480968e-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": 0.01722201881930232, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.008745120884850622, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.04318915829062462, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.03296861108392477, + "weighted_rkd_loss": 0.02159457914531231 + }, + { + "epoch": 0.9731543624161074, + "flow_matching_loss": 0.03220806363970041, + "grad_norm": 0.3510229289531708, + "learning_rate": 2.9646286759550493e-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": 0.017900730203837156, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.008818442607298494, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.044619903340935706, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.03220806363970041, + "weighted_rkd_loss": 0.022309951670467853 + }, + { + "epoch": 0.9899328859060402, + "flow_matching_loss": 0.031706583499908444, + "grad_norm": 0.31205305457115173, + "learning_rate": 2.9624549161589764e-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": 0.019129557255655526, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.010612024832516909, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.04887113943696022, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.031706583499908444, + "weighted_rkd_loss": 0.02443556971848011 + }, + { + "epoch": 1.0067114093959733, + "flow_matching_loss": 0.028854321502149106, + "grad_norm": 0.3063202202320099, + "learning_rate": 2.9602171838471745e-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": 0.022287319228053094, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.010432437527924776, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.05500707514584065, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.028854321502149106, + "weighted_rkd_loss": 0.027503537572920324 + }, + { + "epoch": 1.023489932885906, + "flow_matching_loss": 0.02869299501180649, + "grad_norm": 0.21292831003665924, + "learning_rate": 2.957915576905303e-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": 0.016314297541975976, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.00810289434157312, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.040731489472091195, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.02869299501180649, + "weighted_rkd_loss": 0.020365744736045598 + }, + { + "epoch": 1.0402684563758389, + "flow_matching_loss": 0.026392351277172564, + "grad_norm": 0.315100759267807, + "learning_rate": 2.9555501960131058e-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": 0.01867597484961152, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.009076487878337503, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.046428437903523447, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.026392351277172564, + "weighted_rkd_loss": 0.023214218951761723 + }, + { + "epoch": 1.0570469798657718, + "flow_matching_loss": 0.029286845214664937, + "grad_norm": 0.2971290349960327, + "learning_rate": 2.9531211446400068e-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": 0.01754725184291601, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.009493195638060569, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.04458769857883453, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.029286845214664937, + "weighted_rkd_loss": 0.022293849289417265 + }, + { + "epoch": 1.0738255033557047, + "flow_matching_loss": 0.0279065765440464, + "grad_norm": 0.24345357716083527, + "learning_rate": 2.9506285290405813e-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": 0.016506907157599927, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.008569493470713497, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.041583307832479474, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.0279065765440464, + "weighted_rkd_loss": 0.020791653916239737 + }, + { + "epoch": 1.0906040268456376, + "flow_matching_loss": 0.028139241971075535, + "grad_norm": 0.28294554352760315, + "learning_rate": 2.948072458249911e-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": 0.01751663163304329, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.008426819136366248, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.04346008263528347, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.028139241971075535, + "weighted_rkd_loss": 0.021730041317641734 + }, + { + "epoch": 1.1073825503355705, + "flow_matching_loss": 0.0302914597094059, + "grad_norm": 0.29338598251342773, + "learning_rate": 2.9454530440788125e-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": 0.016945809498429297, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.008697053045034408, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.042588671669363976, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.0302914597094059, + "weighted_rkd_loss": 0.021294335834681988 + }, + { + "epoch": 1.1241610738255035, + "flow_matching_loss": 0.03101203441619873, + "grad_norm": 0.29460427165031433, + "learning_rate": 2.9427704011089474e-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": 0.0163684475235641, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.007941132690757514, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.040678026899695394, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.03101203441619873, + "weighted_rkd_loss": 0.020339013449847697 + }, + { + "epoch": 1.1409395973154361, + "flow_matching_loss": 0.03161245584487915, + "grad_norm": 0.37162813544273376, + "learning_rate": 2.94002464668781e-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": 0.016570307221263646, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.008306508604437112, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.04144712258130312, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.03161245584487915, + "weighted_rkd_loss": 0.02072356129065156 + }, + { + "epoch": 1.157718120805369, + "flow_matching_loss": 0.028309842012822627, + "grad_norm": 0.30125313997268677, + "learning_rate": 2.9372159009235936e-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": 0.01609861133620143, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.008281872188672424, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.040479094721376896, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.028309842012822627, + "weighted_rkd_loss": 0.020239547360688448 + }, + { + "epoch": 1.174496644295302, + "flow_matching_loss": 0.030241508968174458, + "grad_norm": 0.29987332224845886, + "learning_rate": 2.9343442866799365e-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": 0.017228823527693747, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.007661824766546488, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.042119472101330754, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.030241508968174458, + "weighted_rkd_loss": 0.021059736050665377 + }, + { + "epoch": 1.191275167785235, + "flow_matching_loss": 0.027564298920333387, + "grad_norm": 0.33453255891799927, + "learning_rate": 2.9314099295705495e-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": 0.016065255925059318, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.007784039247781038, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.039914550818502904, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.027564298920333387, + "weighted_rkd_loss": 0.019957275409251452 + }, + { + "epoch": 1.2080536912751678, + "flow_matching_loss": 0.026224339753389357, + "grad_norm": 0.3476935029029846, + "learning_rate": 2.928412957953718e-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": 0.015535871591418981, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0071810139808803795, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.03825275730341673, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.026224339753389357, + "weighted_rkd_loss": 0.019126378651708364 + }, + { + "epoch": 1.2248322147651007, + "flow_matching_loss": 0.027642077021300792, + "grad_norm": 0.3133049011230469, + "learning_rate": 2.925353502926689e-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": 0.013876958843320608, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0066064754500985146, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.03436039350926876, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.027642077021300792, + "weighted_rkd_loss": 0.01718019675463438 + }, + { + "epoch": 1.2416107382550337, + "flow_matching_loss": 0.023338249325752257, + "grad_norm": 0.290968656539917, + "learning_rate": 2.922231698319937e-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": 0.014192284736782312, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.006580912135541439, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.03496548105031252, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.023338249325752257, + "weighted_rkd_loss": 0.01748274052515626 + }, + { + "epoch": 1.2583892617449663, + "flow_matching_loss": 0.027370731346309185, + "grad_norm": 0.27969351410865784, + "learning_rate": 2.9190476806913088e-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": 0.016638016607612372, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.00719439540989697, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.04047042895108462, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.027370731346309185, + "weighted_rkd_loss": 0.02023521447554231 + }, + { + "epoch": 1.2751677852348993, + "flow_matching_loss": 0.025700754672288894, + "grad_norm": 0.21908466517925262, + "learning_rate": 2.9158015893200504e-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": 0.015719501581043006, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.007590184360742569, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.03902918752282858, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.025700754672288894, + "weighted_rkd_loss": 0.01951459376141429 + }, + { + "epoch": 1.2919463087248322, + "flow_matching_loss": 0.027718209475278855, + "grad_norm": 0.24782724678516388, + "learning_rate": 2.9124935662007137e-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": 0.015058185812085867, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.007072171056643128, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.037188542820513246, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.027718209475278855, + "weighted_rkd_loss": 0.018594271410256623 + }, + { + "epoch": 1.308724832214765, + "flow_matching_loss": 0.02648232225328684, + "grad_norm": 0.2891945540904999, + "learning_rate": 2.9091237560369467e-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": 0.016410541906952857, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.009014681633561849, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.04183576498180628, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.02648232225328684, + "weighted_rkd_loss": 0.02091788249090314 + }, + { + "epoch": 1.325503355704698, + "flow_matching_loss": 0.02804630771279335, + "grad_norm": 0.2602086663246155, + "learning_rate": 2.905692306235163e-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": 0.014199693780392408, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.006822896050289273, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.03522228356450796, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.02804630771279335, + "weighted_rkd_loss": 0.01761114178225398 + }, + { + "epoch": 1.342281879194631, + "flow_matching_loss": 0.026005209982395174, + "grad_norm": 0.30106857419013977, + "learning_rate": 2.9021993668980925e-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": 0.01492576440796256, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.006834634952247143, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.03668616376817226, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.026005209982395174, + "weighted_rkd_loss": 0.01834308188408613 + }, + { + "epoch": 1.3590604026845639, + "flow_matching_loss": 0.028171230293810367, + "grad_norm": 0.27645203471183777, + "learning_rate": 2.8986450908182185e-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": 0.014832210075110198, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0071496111806482075, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.0368140310049057, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.028171230293810367, + "weighted_rkd_loss": 0.01840701550245285 + }, + { + "epoch": 1.3758389261744965, + "flow_matching_loss": 0.02719732839614153, + "grad_norm": 0.2552866041660309, + "learning_rate": 2.8950296334710902e-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": 0.014208548236638308, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.006532883737236261, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.03494998011738062, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.02719732839614153, + "weighted_rkd_loss": 0.01747499005869031 + }, + { + "epoch": 1.3926174496644295, + "flow_matching_loss": 0.028093678317964076, + "grad_norm": 0.32355931401252747, + "learning_rate": 2.8913531530085248e-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": 0.014492917619645596, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.006872892193496227, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.035858727246522906, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.028093678317964076, + "weighted_rkd_loss": 0.017929363623261453 + }, + { + "epoch": 1.4093959731543624, + "flow_matching_loss": 0.027267585135996343, + "grad_norm": 0.2849968671798706, + "learning_rate": 2.887615810251687e-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": 0.014031628984957934, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.006683602603152395, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.03474686034023762, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.027267585135996343, + "weighted_rkd_loss": 0.01737343017011881 + }, + { + "epoch": 1.4261744966442953, + "flow_matching_loss": 0.02511601522564888, + "grad_norm": 0.32898736000061035, + "learning_rate": 2.8838177686840577e-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": 0.016200020723044873, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.00793383438140154, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.040333876572549346, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.02511601522564888, + "weighted_rkd_loss": 0.020166938286274673 + }, + { + "epoch": 1.4429530201342282, + "flow_matching_loss": 0.026841339468955994, + "grad_norm": 0.2851686477661133, + "learning_rate": 2.8799591944442776e-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": 0.015511635411530732, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.007340135518461466, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.03836340606212616, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.026841339468955994, + "weighted_rkd_loss": 0.01918170303106308 + }, + { + "epoch": 1.4597315436241611, + "flow_matching_loss": 0.02600593287497759, + "grad_norm": 0.3380154073238373, + "learning_rate": 2.8760402563188848e-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": 0.016814271081238986, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.008359821978956461, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.041988364234566686, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.02600593287497759, + "weighted_rkd_loss": 0.020994182117283343 + }, + { + "epoch": 1.476510067114094, + "flow_matching_loss": 0.025509308464825152, + "grad_norm": 0.2603253126144409, + "learning_rate": 2.8720611257349285e-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": 0.013912760373204946, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0063000186812132595, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.0341255396604538, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.025509308464825152, + "weighted_rkd_loss": 0.0170627698302269 + }, + { + "epoch": 1.4932885906040267, + "flow_matching_loss": 0.02826503273099661, + "grad_norm": 0.24795158207416534, + "learning_rate": 2.868021976752471e-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": 0.013900027424097062, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0066476063802838326, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.0344476617872715, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.02826503273099661, + "weighted_rkd_loss": 0.01722383089363575 + }, + { + "epoch": 1.5100671140939599, + "flow_matching_loss": 0.026415893621742726, + "grad_norm": 0.31570133566856384, + "learning_rate": 2.8639229860569738e-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": 0.015136737655848265, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0071802884340286255, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.037453764118254185, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.026415893621742726, + "weighted_rkd_loss": 0.018726882059127092 + }, + { + "epoch": 1.5268456375838926, + "flow_matching_loss": 0.02518075369298458, + "grad_norm": 0.24654585123062134, + "learning_rate": 2.8597643329515688e-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": 0.01396057354286313, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.006632884033024311, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.03455403074622154, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.02518075369298458, + "weighted_rkd_loss": 0.01727701537311077 + }, + { + "epoch": 1.5436241610738255, + "flow_matching_loss": 0.026818909496068955, + "grad_norm": 0.2850186824798584, + "learning_rate": 2.855546199349214e-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": 0.012533736135810613, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.005811271606944501, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.030878743901848794, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.026818909496068955, + "weighted_rkd_loss": 0.015439371950924397 + }, + { + "epoch": 1.5604026845637584, + "flow_matching_loss": 0.02408020496368408, + "grad_norm": 0.3091691732406616, + "learning_rate": 2.85126876976474e-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": 0.012264856789261102, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.00596089877653867, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.030490612238645555, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.02408020496368408, + "weighted_rkd_loss": 0.015245306119322777 + }, + { + "epoch": 1.5771812080536913, + "flow_matching_loss": 0.025282399170100688, + "grad_norm": 0.25739333033561707, + "learning_rate": 2.8469322313067726e-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": 0.015094980597496033, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0074165277415886525, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.037606488727033135, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.025282399170100688, + "weighted_rkd_loss": 0.018803244363516568 + }, + { + "epoch": 1.5939597315436242, + "flow_matching_loss": 0.026252748258411885, + "grad_norm": 0.2987407445907593, + "learning_rate": 2.8425367736695522e-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": 0.011627537570893765, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.005372069217264653, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.028627144172787666, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.026252748258411885, + "weighted_rkd_loss": 0.014313572086393833 + }, + { + "epoch": 1.610738255033557, + "flow_matching_loss": 0.02298819851130247, + "grad_norm": 0.2788366973400116, + "learning_rate": 2.838082589124636e-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": 0.014043763000518083, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.006708415551111102, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.034795941412448884, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.02298819851130247, + "weighted_rkd_loss": 0.017397970706224442 + }, + { + "epoch": 1.62751677852349, + "flow_matching_loss": 0.02329830452799797, + "grad_norm": 0.321382999420166, + "learning_rate": 2.8335698725124843e-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": 0.013656491879373789, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.006472870148718357, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.03378585446625948, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.02329830452799797, + "weighted_rkd_loss": 0.01689292723312974 + }, + { + "epoch": 1.6442953020134228, + "flow_matching_loss": 0.022350782155990602, + "grad_norm": 0.2501571476459503, + "learning_rate": 2.8289988212339422e-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": 0.012651610746979713, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.005716795939952135, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.031020017340779305, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.022350782155990602, + "weighted_rkd_loss": 0.015510008670389652 + }, + { + "epoch": 1.6610738255033557, + "flow_matching_loss": 0.021388859674334525, + "grad_norm": 0.21727903187274933, + "learning_rate": 2.8243696352415997e-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": 0.013748683407902718, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.006664992216974497, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.03416235968470573, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.021388859674334525, + "weighted_rkd_loss": 0.017081179842352866 + }, + { + "epoch": 1.6778523489932886, + "flow_matching_loss": 0.025073782540857792, + "grad_norm": 0.2569715678691864, + "learning_rate": 2.8196825170310495e-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": 0.01149385841563344, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.00539817709941417, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.028385894000530244, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.025073782540857792, + "weighted_rkd_loss": 0.014192947000265122 + } + ], + "logging_steps": 10, + "max_steps": 5000, + "num_input_tokens_seen": 0, + "num_train_epochs": 9, + "save_steps": 5001, + "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": 64, + "trial_name": null, + "trial_params": null +} diff --git a/FewShot/39tasks/5shot/FromRKD-Timewarp_FullactionHead/seed0/default/train_5shot_seed0_fullactionhead_5k/checkpoint-3000/config.json b/FewShot/39tasks/5shot/FromRKD-Timewarp_FullactionHead/seed0/default/train_5shot_seed0_fullactionhead_5k/checkpoint-3000/config.json new file mode 100644 index 0000000000000000000000000000000000000000..4740d0455bef31c469ee5dab6d4a47d64df61665 --- /dev/null +++ b/FewShot/39tasks/5shot/FromRKD-Timewarp_FullactionHead/seed0/default/train_5shot_seed0_fullactionhead_5k/checkpoint-3000/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": 0.5, + "rkd_loss_weight_end": 0.0, + "rkd_loss_weight_schedule": null, + "rkd_loss_weight_start": 0.1, + "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/FewShot/39tasks/5shot/FromRKD-Timewarp_FullactionHead/seed0/default/train_5shot_seed0_fullactionhead_5k/checkpoint-3000/experiment_cfg/metadata.json b/FewShot/39tasks/5shot/FromRKD-Timewarp_FullactionHead/seed0/default/train_5shot_seed0_fullactionhead_5k/checkpoint-3000/experiment_cfg/metadata.json new file mode 100644 index 0000000000000000000000000000000000000000..6f73f8489294c58acc82d1351a17e5e3ff316e18 --- /dev/null +++ b/FewShot/39tasks/5shot/FromRKD-Timewarp_FullactionHead/seed0/default/train_5shot_seed0_fullactionhead_5k/checkpoint-3000/experiment_cfg/metadata.json @@ -0,0 +1,431 @@ +{ + "new_embodiment": { + "statistics": { + "state": { + "base_position": { + "max": [ + 7.778976917266846, + 0.33403685688972473, + 0.7223060727119446 + ], + "min": [ + -4.821934223175049, + -6.785110950469971, + 0.6994727253913879 + ], + "mean": [ + 2.5886094736594543, + -2.1559664577316586, + 0.7003580854548973 + ], + "std": [ + 1.6717149343420803, + 1.3743695399555862, + 0.000940069542359506 + ], + "q01": [ + -0.7517815060249613, + -5.497849509089705, + 0.6999776914424374 + ], + "q99": [ + 6.060945690807028, + -0.5331499932493624, + 0.7028864260472257 + ] + }, + "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, + 0.24488939171811902, + 0.5968549204077488 + ], + "std": [ + 0.0, + 0.0, + 0.6659970734507764, + 0.37443298392262264 + ], + "q01": [ + 0.0, + 0.0, + -0.9999999999999999, + 4.149924358336069e-06 + ], + "q99": [ + 0.0, + 0.0, + 0.9999999999999999, + 0.9999999999999999 + ] + }, + "end_effector_position_relative": { + "max": [ + 0.8848381638526917, + 0.8569459915161133, + 1.053884744644165 + ], + "min": [ + -0.5828549861907959, + -0.8486236333847046, + -0.3446122705936432 + ], + "mean": [ + 0.25087486209693133, + -0.005738920462036393, + 0.43643948866361787 + ], + "std": [ + 0.15741544532518012, + 0.2547603106793105, + 0.2050434492495558 + ], + "q01": [ + -0.013990057288082561, + -0.4824599092623036, + 0.18825220353500022 + ], + "q99": [ + 0.5411989642841469, + 0.4754923657245658, + 0.6912323747882034 + ] + }, + "end_effector_rotation_relative": { + "max": [ + 0.9999949932098389, + 0.9997354745864868, + 0.9900747537612915, + 0.9735639095306396 + ], + "min": [ + -0.9999914169311523, + -0.9999884366989136, + -0.9975369572639465, + 6.754255821306288e-08 + ], + "mean": [ + -0.22290331318740714, + 0.04579037037503981, + -0.05804094906527823, + 0.12347219024917105 + ], + "std": [ + 0.8519267931761344, + 0.2949394732270141, + 0.30026078832556913, + 0.16333086331829588 + ], + "q01": [ + -0.9957959450418518, + -0.5678869493126506, + -0.5318135475761875, + 0.0014160465686576505 + ], + "q99": [ + 0.9900692296898317, + 0.6357358195673684, + 0.5009847868015971, + 0.4287044482210845 + ] + }, + "gripper_qpos": { + "max": [ + 0.05289580300450325, + 0.00937709677964449 + ], + "min": [ + -0.011669117026031017, + -0.06307756155729294 + ], + "mean": [ + 0.03203268930378034, + -0.032331639271968075 + ], + "std": [ + 0.011607049173153697, + 0.011306286745872529 + ], + "q01": [ + 0.006429109393656672, + -0.04056586943958638 + ], + "q99": [ + 0.040377860142451545, + -0.006593889698673258 + ] + } + }, + "action": { + "base_motion": { + "max": [ + 1.0, + 1.0, + 1.0, + 0.0 + ], + "min": [ + -1.0, + -1.0, + -1.0, + 0.0 + ], + "mean": [ + 0.006050336852760097, + 0.002728916466628117, + -0.0017236811393598838, + 0.0 + ], + "std": [ + 0.11032167711449951, + 0.09824140301615687, + 0.07825208047515658, + 0.0 + ], + "q01": [ + -0.154721262094005, + -0.18955302988697728, + -0.14454065867754942, + 0.0 + ], + "q99": [ + 0.24492955247744197, + 0.2496797822190429, + 0.11966742325336034, + 0.0 + ] + }, + "control_mode": { + "max": [ + 1.0 + ], + "min": [ + -1.0 + ], + "mean": [ + -0.9294448715752568 + ], + "std": [ + 0.36897903503358326 + ], + "q01": [ + -1.0 + ], + "q99": [ + -0.21627562513311385 + ] + }, + "end_effector_position": { + "max": [ + 1.0, + 1.0, + 1.0 + ], + "min": [ + -1.0, + -1.0, + -1.0 + ], + "mean": [ + -0.00826780165926883, + 0.003161997359379232, + -0.08157453651841007 + ], + "std": [ + 0.4545430944464769, + 0.41040151434986594, + 0.4102644251536544 + ], + "q01": [ + -0.9653185255999107, + -0.9255635199318754, + -0.8208437313341718 + ], + "q99": [ + 0.7987781084592441, + 0.9260844232330088, + 0.8357421812622201 + ] + }, + "end_effector_rotation": { + "max": [ + 1.0, + 1.0, + 1.0 + ], + "min": [ + -1.0, + -1.0, + -1.0 + ], + "mean": [ + 0.007367024641295137, + -0.02060756343728361, + -0.001884833075998489 + ], + "std": [ + 0.1116768956941076, + 0.13249042272165212, + 0.12338908364725322 + ], + "q01": [ + -0.26894658663246845, + -0.39865607740339326, + -0.34393735191302655 + ], + "q99": [ + 0.3191674780171587, + 0.31133630397431267, + 0.33968476423051097 + ] + }, + "gripper_close": { + "max": [ + 1.0 + ], + "min": [ + -1.0 + ], + "mean": [ + -0.3098994894722959 + ], + "std": [ + 0.9507718287498876 + ], + "q01": [ + -1.0 + ], + "q99": [ + 0.7545620899936146 + ] + } + } + }, + "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/FewShot/39tasks/5shot/FromRKD-Timewarp_FullactionHead/seed0/default/train_5shot_seed0_fullactionhead_5k/checkpoint-3000/model.safetensors.index.json b/FewShot/39tasks/5shot/FromRKD-Timewarp_FullactionHead/seed0/default/train_5shot_seed0_fullactionhead_5k/checkpoint-3000/model.safetensors.index.json new file mode 100644 index 0000000000000000000000000000000000000000..b20baed71f4d16a411223b70d45d934c6cbbfe97 --- /dev/null +++ b/FewShot/39tasks/5shot/FromRKD-Timewarp_FullactionHead/seed0/default/train_5shot_seed0_fullactionhead_5k/checkpoint-3000/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/FewShot/39tasks/5shot/FromRKD-Timewarp_FullactionHead/seed0/default/train_5shot_seed0_fullactionhead_5k/checkpoint-3000/trainer_state.json b/FewShot/39tasks/5shot/FromRKD-Timewarp_FullactionHead/seed0/default/train_5shot_seed0_fullactionhead_5k/checkpoint-3000/trainer_state.json new file mode 100644 index 0000000000000000000000000000000000000000..6a4450469a39f216e1fcb1710c3b382487069cca --- /dev/null +++ b/FewShot/39tasks/5shot/FromRKD-Timewarp_FullactionHead/seed0/default/train_5shot_seed0_fullactionhead_5k/checkpoint-3000/trainer_state.json @@ -0,0 +1,13834 @@ +{ + "best_global_step": null, + "best_metric": null, + "best_model_checkpoint": null, + "epoch": 5.033557046979865, + "eval_steps": 500, + "global_step": 3000, + "is_hyper_param_search": false, + "is_local_process_zero": true, + "is_world_process_zero": true, + "log_history": [ + { + "epoch": 0.016778523489932886, + "flow_matching_loss": 0.189423368871212, + "grad_norm": 1.1502586603164673, + "learning_rate": 1.08e-06, + "loss": 0.38, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.13627509251236916, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.10864389836788177, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.38119408786296843, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.189423368871212, + "weighted_rkd_loss": 0.19059704393148422 + }, + { + "epoch": 0.03355704697986577, + "flow_matching_loss": 0.21835082918405532, + "grad_norm": 1.3782790899276733, + "learning_rate": 2.28e-06, + "loss": 0.4112, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.13762855604290963, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.11038857251405716, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.3856456845998764, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.21835082918405532, + "weighted_rkd_loss": 0.1928228422999382 + }, + { + "epoch": 0.050335570469798654, + "flow_matching_loss": 0.1594226397573948, + "grad_norm": 1.0711982250213623, + "learning_rate": 3.48e-06, + "loss": 0.3382, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.1283317506313324, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.10086342692375183, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.3575269281864166, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.1594226397573948, + "weighted_rkd_loss": 0.1787634640932083 + }, + { + "epoch": 0.06711409395973154, + "flow_matching_loss": 0.15615404024720192, + "grad_norm": 0.7945505380630493, + "learning_rate": 4.68e-06, + "loss": 0.3226, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.12152766585350036, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0897677019238472, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.33282303512096406, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.15615404024720192, + "weighted_rkd_loss": 0.16641151756048203 + }, + { + "epoch": 0.08389261744966443, + "flow_matching_loss": 0.13071623146533967, + "grad_norm": 0.8744078278541565, + "learning_rate": 5.8800000000000005e-06, + "loss": 0.2952, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.11974534913897514, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.08949827328324318, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.32898896336555483, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.13071623146533967, + "weighted_rkd_loss": 0.16449448168277742 + }, + { + "epoch": 0.10067114093959731, + "flow_matching_loss": 0.11403126865625382, + "grad_norm": 0.651993453502655, + "learning_rate": 7.08e-06, + "loss": 0.2721, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.11685370355844497, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.08247915208339691, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.3161865621805191, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.11403126865625382, + "weighted_rkd_loss": 0.15809328109025955 + }, + { + "epoch": 0.1174496644295302, + "flow_matching_loss": 0.0997005932033062, + "grad_norm": 0.6273358464241028, + "learning_rate": 8.28e-06, + "loss": 0.2453, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.10773090794682502, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.07566531300544739, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.29112712740898133, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.0997005932033062, + "weighted_rkd_loss": 0.14556356370449067 + }, + { + "epoch": 0.1342281879194631, + "flow_matching_loss": 0.08913109861314297, + "grad_norm": 0.6481125950813293, + "learning_rate": 9.48e-06, + "loss": 0.2235, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.10180731341242791, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.06521718949079514, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.2688318148255348, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.08913109861314297, + "weighted_rkd_loss": 0.1344159074127674 + }, + { + "epoch": 0.15100671140939598, + "flow_matching_loss": 0.07915030717849732, + "grad_norm": 0.6114252209663391, + "learning_rate": 1.068e-05, + "loss": 0.2091, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.09674249440431595, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.06643815003335476, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.2599231392145157, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.07915030717849732, + "weighted_rkd_loss": 0.12996156960725785 + }, + { + "epoch": 0.16778523489932887, + "flow_matching_loss": 0.07286151051521302, + "grad_norm": 0.5393626093864441, + "learning_rate": 1.1880000000000001e-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": 0.08473351448774338, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.054971330985426906, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.2244383603334427, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.07286151051521302, + "weighted_rkd_loss": 0.11221918016672135 + }, + { + "epoch": 0.18456375838926176, + "flow_matching_loss": 0.07065574415028095, + "grad_norm": 0.6259891390800476, + "learning_rate": 1.308e-05, + "loss": 0.1874, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.08815310075879097, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.05715718045830727, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.23346338123083116, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.07065574415028095, + "weighted_rkd_loss": 0.11673169061541558 + }, + { + "epoch": 0.20134228187919462, + "flow_matching_loss": 0.0647046722471714, + "grad_norm": 0.7180153131484985, + "learning_rate": 1.428e-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": 0.07692675888538361, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.05016751065850258, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.20402102917432785, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.0647046722471714, + "weighted_rkd_loss": 0.10201051458716393 + }, + { + "epoch": 0.2181208053691275, + "flow_matching_loss": 0.06340371444821358, + "grad_norm": 0.5939756035804749, + "learning_rate": 1.548e-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": 0.07837383523583412, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.04749012887477875, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.20423779934644698, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.06340371444821358, + "weighted_rkd_loss": 0.10211889967322349 + }, + { + "epoch": 0.2348993288590604, + "flow_matching_loss": 0.06330732218921184, + "grad_norm": 0.5114864110946655, + "learning_rate": 1.6680000000000003e-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": 0.07281922809779644, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.04322611875832081, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.1888645753264427, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.06330732218921184, + "weighted_rkd_loss": 0.09443228766322136 + }, + { + "epoch": 0.2516778523489933, + "flow_matching_loss": 0.05951428152620793, + "grad_norm": 0.6290663480758667, + "learning_rate": 1.7879999999999998e-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": 0.06923839673399926, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.041895052045583726, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.18037184625864028, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.05951428152620793, + "weighted_rkd_loss": 0.09018592312932014 + }, + { + "epoch": 0.2684563758389262, + "flow_matching_loss": 0.058728757128119466, + "grad_norm": 0.5164991021156311, + "learning_rate": 1.908e-05, + "loss": 0.1374, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.060720610246062276, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0359353007748723, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.15737652033567429, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.058728757128119466, + "weighted_rkd_loss": 0.07868826016783714 + }, + { + "epoch": 0.28523489932885904, + "flow_matching_loss": 0.05690980330109596, + "grad_norm": 0.42798617482185364, + "learning_rate": 2.0280000000000002e-05, + "loss": 0.1344, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.05981909520924091, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.03533888217061758, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.1549770712852478, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.05690980330109596, + "weighted_rkd_loss": 0.0774885356426239 + }, + { + "epoch": 0.30201342281879195, + "flow_matching_loss": 0.05503378733992577, + "grad_norm": 0.4299779534339905, + "learning_rate": 2.148e-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": 0.05938076414167881, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.03561958894133568, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.15438111796975135, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.05503378733992577, + "weighted_rkd_loss": 0.07719055898487567 + }, + { + "epoch": 0.3187919463087248, + "flow_matching_loss": 0.054644962027668956, + "grad_norm": 0.4563824534416199, + "learning_rate": 2.268e-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": 0.05716169141232967, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.03335041459649801, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.147673799097538, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.054644962027668956, + "weighted_rkd_loss": 0.073836899548769 + }, + { + "epoch": 0.33557046979865773, + "flow_matching_loss": 0.056772416457533836, + "grad_norm": 0.4835961163043976, + "learning_rate": 2.3880000000000002e-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": 0.054468106105923654, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.03251056261360645, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.14144677594304084, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.056772416457533836, + "weighted_rkd_loss": 0.07072338797152042 + }, + { + "epoch": 0.3523489932885906, + "flow_matching_loss": 0.0495440598577261, + "grad_norm": 0.5092470049858093, + "learning_rate": 2.508e-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": 0.04748903773725033, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.026358308084309102, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.12133638188242912, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.0495440598577261, + "weighted_rkd_loss": 0.06066819094121456 + }, + { + "epoch": 0.3691275167785235, + "flow_matching_loss": 0.0511940747499466, + "grad_norm": 0.4401130974292755, + "learning_rate": 2.628e-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": 0.0482368066906929, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0273994205519557, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.1238730326294899, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.0511940747499466, + "weighted_rkd_loss": 0.06193651631474495 + }, + { + "epoch": 0.3859060402684564, + "flow_matching_loss": 0.050138309597969055, + "grad_norm": 0.42907798290252686, + "learning_rate": 2.748e-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": 0.0430266585201025, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.025129231996834277, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.11118254959583282, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.050138309597969055, + "weighted_rkd_loss": 0.05559127479791641 + }, + { + "epoch": 0.40268456375838924, + "flow_matching_loss": 0.046704120188951495, + "grad_norm": 0.4718972146511078, + "learning_rate": 2.868e-05, + "loss": 0.1016, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.042805034294724466, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.024135740101337434, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.10974581018090249, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.046704120188951495, + "weighted_rkd_loss": 0.05487290509045124 + }, + { + "epoch": 0.41946308724832215, + "flow_matching_loss": 0.044243189692497256, + "grad_norm": 0.3790624737739563, + "learning_rate": 2.9880000000000002e-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": 0.041216658987104894, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.021118617989122866, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.10355193763971329, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.044243189692497256, + "weighted_rkd_loss": 0.051775968819856646 + }, + { + "epoch": 0.436241610738255, + "flow_matching_loss": 0.04542431682348251, + "grad_norm": 0.512637197971344, + "learning_rate": 2.9999734259635676e-05, + "loss": 0.0959, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.03973671272397041, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.021489275991916655, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.10096270069479943, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.04542431682348251, + "weighted_rkd_loss": 0.050481350347399714 + }, + { + "epoch": 0.45302013422818793, + "flow_matching_loss": 0.05049298815429211, + "grad_norm": 0.4598679840564728, + "learning_rate": 2.9998815663057245e-05, + "loss": 0.102, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.040620113164186476, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.021747449319809674, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.10298767462372779, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.05049298815429211, + "weighted_rkd_loss": 0.051493837311863896 + }, + { + "epoch": 0.4697986577181208, + "flow_matching_loss": 0.046248992905020714, + "grad_norm": 0.4032992422580719, + "learning_rate": 2.999724096969228e-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": 0.038419022038578984, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.021234015561640264, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.09807205945253372, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.046248992905020714, + "weighted_rkd_loss": 0.04903602972626686 + }, + { + "epoch": 0.4865771812080537, + "flow_matching_loss": 0.04581176079809666, + "grad_norm": 0.378512978553772, + "learning_rate": 2.9995010248422984e-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": 0.036053233221173285, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.019097956269979476, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.09120442345738411, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.04581176079809666, + "weighted_rkd_loss": 0.045602211728692055 + }, + { + "epoch": 0.5033557046979866, + "flow_matching_loss": 0.040904562920331955, + "grad_norm": 0.3660767078399658, + "learning_rate": 2.999212359682832e-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": 0.030510597676038743, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.015882878191769124, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.07690407261252404, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.040904562920331955, + "weighted_rkd_loss": 0.03845203630626202 + }, + { + "epoch": 0.5201342281879194, + "flow_matching_loss": 0.03941964264959097, + "grad_norm": 0.3793014883995056, + "learning_rate": 2.9988581141179787e-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": 0.032410606555640695, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.01629540789872408, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.08111662119626999, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.03941964264959097, + "weighted_rkd_loss": 0.04055831059813499 + }, + { + "epoch": 0.5369127516778524, + "flow_matching_loss": 0.04001129996031523, + "grad_norm": 0.6170795559883118, + "learning_rate": 2.998438303643587e-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": 0.033591178432106975, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.018490686919540165, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.08567304462194443, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.04001129996031523, + "weighted_rkd_loss": 0.042836522310972215 + }, + { + "epoch": 0.5536912751677853, + "flow_matching_loss": 0.040321177989244464, + "grad_norm": 0.3570716381072998, + "learning_rate": 2.997952946623527e-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": 0.031730389967560765, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.016793423146009446, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.08025420233607292, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.040321177989244464, + "weighted_rkd_loss": 0.04012710116803646 + }, + { + "epoch": 0.5704697986577181, + "flow_matching_loss": 0.04315701965242624, + "grad_norm": 0.379740446805954, + "learning_rate": 2.9974020642888875e-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": 0.03032686710357666, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.01704806601628661, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.07770180031657219, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.04315701965242624, + "weighted_rkd_loss": 0.03885090015828609 + }, + { + "epoch": 0.587248322147651, + "flow_matching_loss": 0.040681138448417184, + "grad_norm": 0.33822014927864075, + "learning_rate": 2.9967856807370458e-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": 0.027930348366498946, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0157440728507936, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.0716047689318657, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.040681138448417184, + "weighted_rkd_loss": 0.03580238446593285 + }, + { + "epoch": 0.6040268456375839, + "flow_matching_loss": 0.042508364841341975, + "grad_norm": 0.40825992822647095, + "learning_rate": 2.996103822930615e-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": 0.031780967675149444, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.016943857539445162, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.08050579428672791, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.042508364841341975, + "weighted_rkd_loss": 0.040252897143363955 + }, + { + "epoch": 0.6208053691275168, + "flow_matching_loss": 0.037045056186616424, + "grad_norm": 0.3486769199371338, + "learning_rate": 2.9953565206962653e-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": 0.028993065282702447, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.014874111022800207, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.07286024205386639, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.037045056186616424, + "weighted_rkd_loss": 0.036430121026933195 + }, + { + "epoch": 0.6375838926174496, + "flow_matching_loss": 0.038407695665955546, + "grad_norm": 0.4263972342014313, + "learning_rate": 2.9945438067234172e-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": 0.029730910249054433, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.015382550936192275, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.07484437264502049, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.038407695665955546, + "weighted_rkd_loss": 0.037422186322510244 + }, + { + "epoch": 0.6543624161073825, + "flow_matching_loss": 0.038366636447608474, + "grad_norm": 0.3533881604671478, + "learning_rate": 2.9936657165628124e-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": 0.027479714527726174, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.014265698939561844, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.0692251279950142, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.038366636447608474, + "weighted_rkd_loss": 0.0346125639975071 + }, + { + "epoch": 0.6711409395973155, + "flow_matching_loss": 0.035504350997507574, + "grad_norm": 0.3166177272796631, + "learning_rate": 2.992722288624959e-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": 0.025187848880887032, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.013730826880782842, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.06410652436316014, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.035504350997507574, + "weighted_rkd_loss": 0.03205326218158007 + }, + { + "epoch": 0.6879194630872483, + "flow_matching_loss": 0.037043062038719654, + "grad_norm": 0.3390388488769531, + "learning_rate": 2.9917135641784515e-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": 0.0239717124029994, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.012146859616041183, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.060090285167098044, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.037043062038719654, + "weighted_rkd_loss": 0.030045142583549022 + }, + { + "epoch": 0.7046979865771812, + "flow_matching_loss": 0.03493194635957479, + "grad_norm": 0.3506368398666382, + "learning_rate": 2.9906395873481655e-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": 0.025168376229703426, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.012817893270403146, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.06315464563667775, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.03493194635957479, + "weighted_rkd_loss": 0.031577322818338874 + }, + { + "epoch": 0.7214765100671141, + "flow_matching_loss": 0.036136615090072155, + "grad_norm": 0.39711323380470276, + "learning_rate": 2.989500405113326e-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": 0.025528428331017494, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.012982557248324155, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.06403941474854946, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.036136615090072155, + "weighted_rkd_loss": 0.03201970737427473 + }, + { + "epoch": 0.738255033557047, + "flow_matching_loss": 0.0340481685474515, + "grad_norm": 0.3472777009010315, + "learning_rate": 2.9882960673054547e-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": 0.024420114979147912, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.012645702250301838, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.061485931649804114, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.0340481685474515, + "weighted_rkd_loss": 0.030742965824902057 + }, + { + "epoch": 0.7550335570469798, + "flow_matching_loss": 0.035830936022102834, + "grad_norm": 0.34199073910713196, + "learning_rate": 2.9870266266061886e-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": 0.02066250964999199, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.01092346548102796, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.05224848538637161, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.035830936022102834, + "weighted_rkd_loss": 0.026124242693185806 + }, + { + "epoch": 0.7718120805369127, + "flow_matching_loss": 0.03825925439596176, + "grad_norm": 0.3056429326534271, + "learning_rate": 2.9856921385449766e-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": 0.026111494936048984, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.01336317528039217, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.06558616384863854, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.03825925439596176, + "weighted_rkd_loss": 0.03279308192431927 + }, + { + "epoch": 0.7885906040268457, + "flow_matching_loss": 0.03638801425695419, + "grad_norm": 0.3128032982349396, + "learning_rate": 2.9842926614966502e-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": 0.02268179152160883, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.011993827018886804, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.05735741071403026, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.03638801425695419, + "weighted_rkd_loss": 0.02867870535701513 + }, + { + "epoch": 0.8053691275167785, + "flow_matching_loss": 0.03183348029851914, + "grad_norm": 0.2547442317008972, + "learning_rate": 2.982828256678868e-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": 0.022423543594777585, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.011085591791197658, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.05593267902731895, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.03183348029851914, + "weighted_rkd_loss": 0.027966339513659476 + }, + { + "epoch": 0.8221476510067114, + "flow_matching_loss": 0.03268085308372974, + "grad_norm": 0.2893790006637573, + "learning_rate": 2.9812989881494412e-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": 0.02247578836977482, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.011517873685806989, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.05646945089101792, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.03268085308372974, + "weighted_rkd_loss": 0.02823472544550896 + }, + { + "epoch": 0.8389261744966443, + "flow_matching_loss": 0.038189934380352494, + "grad_norm": 0.4075469672679901, + "learning_rate": 2.9797049228035296e-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": 0.02296894444152713, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.012043415708467365, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.05798130407929421, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.038189934380352494, + "weighted_rkd_loss": 0.028990652039647104 + }, + { + "epoch": 0.8557046979865772, + "flow_matching_loss": 0.03394890353083611, + "grad_norm": 0.36510178446769714, + "learning_rate": 2.978046130370715e-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": 0.02077607475221157, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.01056032245978713, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.0521124716848135, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.03394890353083611, + "weighted_rkd_loss": 0.02605623584240675 + }, + { + "epoch": 0.87248322147651, + "flow_matching_loss": 0.03357741795480251, + "grad_norm": 0.2992410361766815, + "learning_rate": 2.9763226834119535e-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": 0.020533321145921945, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.010199433751404285, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.05126607567071915, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.03357741795480251, + "weighted_rkd_loss": 0.025633037835359574 + }, + { + "epoch": 0.889261744966443, + "flow_matching_loss": 0.03181673102080822, + "grad_norm": 0.30052074790000916, + "learning_rate": 2.9745346573163963e-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": 0.018613563012331726, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.009025087859481574, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.04625221379101276, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.03181673102080822, + "weighted_rkd_loss": 0.02312610689550638 + }, + { + "epoch": 0.9060402684563759, + "flow_matching_loss": 0.031556636840105054, + "grad_norm": 0.28555840253829956, + "learning_rate": 2.9726821302980987e-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": 0.018975557200610636, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.010126672731712461, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.048077787086367604, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.031556636840105054, + "weighted_rkd_loss": 0.024038893543183802 + }, + { + "epoch": 0.9228187919463087, + "flow_matching_loss": 0.031900867260992526, + "grad_norm": 0.3493785560131073, + "learning_rate": 2.9707651833925935e-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": 0.01944209486246109, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.00898348679766059, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.047867676615715025, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.031900867260992526, + "weighted_rkd_loss": 0.023933838307857513 + }, + { + "epoch": 0.9395973154362416, + "flow_matching_loss": 0.033641642332077025, + "grad_norm": 0.3126336932182312, + "learning_rate": 2.968783900453348e-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": 0.018097881507128478, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.009972660942003131, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.04616842418909073, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.033641642332077025, + "weighted_rkd_loss": 0.023084212094545364 + }, + { + "epoch": 0.9563758389261745, + "flow_matching_loss": 0.03296861108392477, + "grad_norm": 0.3514993488788605, + "learning_rate": 2.9667383681480968e-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": 0.01722201881930232, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.008745120884850622, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.04318915829062462, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.03296861108392477, + "weighted_rkd_loss": 0.02159457914531231 + }, + { + "epoch": 0.9731543624161074, + "flow_matching_loss": 0.03220806363970041, + "grad_norm": 0.3510229289531708, + "learning_rate": 2.9646286759550493e-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": 0.017900730203837156, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.008818442607298494, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.044619903340935706, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.03220806363970041, + "weighted_rkd_loss": 0.022309951670467853 + }, + { + "epoch": 0.9899328859060402, + "flow_matching_loss": 0.031706583499908444, + "grad_norm": 0.31205305457115173, + "learning_rate": 2.9624549161589764e-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": 0.019129557255655526, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.010612024832516909, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.04887113943696022, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.031706583499908444, + "weighted_rkd_loss": 0.02443556971848011 + }, + { + "epoch": 1.0067114093959733, + "flow_matching_loss": 0.028854321502149106, + "grad_norm": 0.3063202202320099, + "learning_rate": 2.9602171838471745e-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": 0.022287319228053094, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.010432437527924776, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.05500707514584065, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.028854321502149106, + "weighted_rkd_loss": 0.027503537572920324 + }, + { + "epoch": 1.023489932885906, + "flow_matching_loss": 0.02869299501180649, + "grad_norm": 0.21292831003665924, + "learning_rate": 2.957915576905303e-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": 0.016314297541975976, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.00810289434157312, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.040731489472091195, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.02869299501180649, + "weighted_rkd_loss": 0.020365744736045598 + }, + { + "epoch": 1.0402684563758389, + "flow_matching_loss": 0.026392351277172564, + "grad_norm": 0.315100759267807, + "learning_rate": 2.9555501960131058e-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": 0.01867597484961152, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.009076487878337503, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.046428437903523447, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.026392351277172564, + "weighted_rkd_loss": 0.023214218951761723 + }, + { + "epoch": 1.0570469798657718, + "flow_matching_loss": 0.029286845214664937, + "grad_norm": 0.2971290349960327, + "learning_rate": 2.9531211446400068e-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": 0.01754725184291601, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.009493195638060569, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.04458769857883453, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.029286845214664937, + "weighted_rkd_loss": 0.022293849289417265 + }, + { + "epoch": 1.0738255033557047, + "flow_matching_loss": 0.0279065765440464, + "grad_norm": 0.24345357716083527, + "learning_rate": 2.9506285290405813e-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": 0.016506907157599927, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.008569493470713497, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.041583307832479474, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.0279065765440464, + "weighted_rkd_loss": 0.020791653916239737 + }, + { + "epoch": 1.0906040268456376, + "flow_matching_loss": 0.028139241971075535, + "grad_norm": 0.28294554352760315, + "learning_rate": 2.948072458249911e-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": 0.01751663163304329, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.008426819136366248, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.04346008263528347, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.028139241971075535, + "weighted_rkd_loss": 0.021730041317641734 + }, + { + "epoch": 1.1073825503355705, + "flow_matching_loss": 0.0302914597094059, + "grad_norm": 0.29338598251342773, + "learning_rate": 2.9454530440788125e-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": 0.016945809498429297, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.008697053045034408, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.042588671669363976, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.0302914597094059, + "weighted_rkd_loss": 0.021294335834681988 + }, + { + "epoch": 1.1241610738255035, + "flow_matching_loss": 0.03101203441619873, + "grad_norm": 0.29460427165031433, + "learning_rate": 2.9427704011089474e-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": 0.0163684475235641, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.007941132690757514, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.040678026899695394, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.03101203441619873, + "weighted_rkd_loss": 0.020339013449847697 + }, + { + "epoch": 1.1409395973154361, + "flow_matching_loss": 0.03161245584487915, + "grad_norm": 0.37162813544273376, + "learning_rate": 2.94002464668781e-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": 0.016570307221263646, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.008306508604437112, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.04144712258130312, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.03161245584487915, + "weighted_rkd_loss": 0.02072356129065156 + }, + { + "epoch": 1.157718120805369, + "flow_matching_loss": 0.028309842012822627, + "grad_norm": 0.30125313997268677, + "learning_rate": 2.9372159009235936e-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": 0.01609861133620143, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.008281872188672424, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.040479094721376896, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.028309842012822627, + "weighted_rkd_loss": 0.020239547360688448 + }, + { + "epoch": 1.174496644295302, + "flow_matching_loss": 0.030241508968174458, + "grad_norm": 0.29987332224845886, + "learning_rate": 2.9343442866799365e-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": 0.017228823527693747, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.007661824766546488, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.042119472101330754, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.030241508968174458, + "weighted_rkd_loss": 0.021059736050665377 + }, + { + "epoch": 1.191275167785235, + "flow_matching_loss": 0.027564298920333387, + "grad_norm": 0.33453255891799927, + "learning_rate": 2.9314099295705495e-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": 0.016065255925059318, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.007784039247781038, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.039914550818502904, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.027564298920333387, + "weighted_rkd_loss": 0.019957275409251452 + }, + { + "epoch": 1.2080536912751678, + "flow_matching_loss": 0.026224339753389357, + "grad_norm": 0.3476935029029846, + "learning_rate": 2.928412957953718e-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": 0.015535871591418981, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0071810139808803795, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.03825275730341673, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.026224339753389357, + "weighted_rkd_loss": 0.019126378651708364 + }, + { + "epoch": 1.2248322147651007, + "flow_matching_loss": 0.027642077021300792, + "grad_norm": 0.3133049011230469, + "learning_rate": 2.925353502926689e-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": 0.013876958843320608, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0066064754500985146, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.03436039350926876, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.027642077021300792, + "weighted_rkd_loss": 0.01718019675463438 + }, + { + "epoch": 1.2416107382550337, + "flow_matching_loss": 0.023338249325752257, + "grad_norm": 0.290968656539917, + "learning_rate": 2.922231698319937e-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": 0.014192284736782312, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.006580912135541439, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.03496548105031252, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.023338249325752257, + "weighted_rkd_loss": 0.01748274052515626 + }, + { + "epoch": 1.2583892617449663, + "flow_matching_loss": 0.027370731346309185, + "grad_norm": 0.27969351410865784, + "learning_rate": 2.9190476806913088e-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": 0.016638016607612372, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.00719439540989697, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.04047042895108462, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.027370731346309185, + "weighted_rkd_loss": 0.02023521447554231 + }, + { + "epoch": 1.2751677852348993, + "flow_matching_loss": 0.025700754672288894, + "grad_norm": 0.21908466517925262, + "learning_rate": 2.9158015893200504e-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": 0.015719501581043006, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.007590184360742569, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.03902918752282858, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.025700754672288894, + "weighted_rkd_loss": 0.01951459376141429 + }, + { + "epoch": 1.2919463087248322, + "flow_matching_loss": 0.027718209475278855, + "grad_norm": 0.24782724678516388, + "learning_rate": 2.9124935662007137e-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": 0.015058185812085867, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.007072171056643128, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.037188542820513246, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.027718209475278855, + "weighted_rkd_loss": 0.018594271410256623 + }, + { + "epoch": 1.308724832214765, + "flow_matching_loss": 0.02648232225328684, + "grad_norm": 0.2891945540904999, + "learning_rate": 2.9091237560369467e-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": 0.016410541906952857, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.009014681633561849, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.04183576498180628, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.02648232225328684, + "weighted_rkd_loss": 0.02091788249090314 + }, + { + "epoch": 1.325503355704698, + "flow_matching_loss": 0.02804630771279335, + "grad_norm": 0.2602086663246155, + "learning_rate": 2.905692306235163e-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": 0.014199693780392408, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.006822896050289273, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.03522228356450796, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.02804630771279335, + "weighted_rkd_loss": 0.01761114178225398 + }, + { + "epoch": 1.342281879194631, + "flow_matching_loss": 0.026005209982395174, + "grad_norm": 0.30106857419013977, + "learning_rate": 2.9021993668980925e-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": 0.01492576440796256, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.006834634952247143, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.03668616376817226, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.026005209982395174, + "weighted_rkd_loss": 0.01834308188408613 + }, + { + "epoch": 1.3590604026845639, + "flow_matching_loss": 0.028171230293810367, + "grad_norm": 0.27645203471183777, + "learning_rate": 2.8986450908182185e-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": 0.014832210075110198, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0071496111806482075, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.0368140310049057, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.028171230293810367, + "weighted_rkd_loss": 0.01840701550245285 + }, + { + "epoch": 1.3758389261744965, + "flow_matching_loss": 0.02719732839614153, + "grad_norm": 0.2552866041660309, + "learning_rate": 2.8950296334710902e-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": 0.014208548236638308, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.006532883737236261, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.03494998011738062, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.02719732839614153, + "weighted_rkd_loss": 0.01747499005869031 + }, + { + "epoch": 1.3926174496644295, + "flow_matching_loss": 0.028093678317964076, + "grad_norm": 0.32355931401252747, + "learning_rate": 2.8913531530085248e-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": 0.014492917619645596, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.006872892193496227, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.035858727246522906, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.028093678317964076, + "weighted_rkd_loss": 0.017929363623261453 + }, + { + "epoch": 1.4093959731543624, + "flow_matching_loss": 0.027267585135996343, + "grad_norm": 0.2849968671798706, + "learning_rate": 2.887615810251687e-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": 0.014031628984957934, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.006683602603152395, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.03474686034023762, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.027267585135996343, + "weighted_rkd_loss": 0.01737343017011881 + }, + { + "epoch": 1.4261744966442953, + "flow_matching_loss": 0.02511601522564888, + "grad_norm": 0.32898736000061035, + "learning_rate": 2.8838177686840577e-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": 0.016200020723044873, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.00793383438140154, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.040333876572549346, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.02511601522564888, + "weighted_rkd_loss": 0.020166938286274673 + }, + { + "epoch": 1.4429530201342282, + "flow_matching_loss": 0.026841339468955994, + "grad_norm": 0.2851686477661133, + "learning_rate": 2.8799591944442776e-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": 0.015511635411530732, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.007340135518461466, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.03836340606212616, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.026841339468955994, + "weighted_rkd_loss": 0.01918170303106308 + }, + { + "epoch": 1.4597315436241611, + "flow_matching_loss": 0.02600593287497759, + "grad_norm": 0.3380154073238373, + "learning_rate": 2.8760402563188848e-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": 0.016814271081238986, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.008359821978956461, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.041988364234566686, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.02600593287497759, + "weighted_rkd_loss": 0.020994182117283343 + }, + { + "epoch": 1.476510067114094, + "flow_matching_loss": 0.025509308464825152, + "grad_norm": 0.2603253126144409, + "learning_rate": 2.8720611257349285e-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": 0.013912760373204946, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0063000186812132595, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.0341255396604538, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.025509308464825152, + "weighted_rkd_loss": 0.0170627698302269 + }, + { + "epoch": 1.4932885906040267, + "flow_matching_loss": 0.02826503273099661, + "grad_norm": 0.24795158207416534, + "learning_rate": 2.868021976752471e-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": 0.013900027424097062, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0066476063802838326, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.0344476617872715, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.02826503273099661, + "weighted_rkd_loss": 0.01722383089363575 + }, + { + "epoch": 1.5100671140939599, + "flow_matching_loss": 0.026415893621742726, + "grad_norm": 0.31570133566856384, + "learning_rate": 2.8639229860569738e-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": 0.015136737655848265, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0071802884340286255, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.037453764118254185, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.026415893621742726, + "weighted_rkd_loss": 0.018726882059127092 + }, + { + "epoch": 1.5268456375838926, + "flow_matching_loss": 0.02518075369298458, + "grad_norm": 0.24654585123062134, + "learning_rate": 2.8597643329515688e-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": 0.01396057354286313, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.006632884033024311, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.03455403074622154, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.02518075369298458, + "weighted_rkd_loss": 0.01727701537311077 + }, + { + "epoch": 1.5436241610738255, + "flow_matching_loss": 0.026818909496068955, + "grad_norm": 0.2850186824798584, + "learning_rate": 2.855546199349214e-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": 0.012533736135810613, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.005811271606944501, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.030878743901848794, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.026818909496068955, + "weighted_rkd_loss": 0.015439371950924397 + }, + { + "epoch": 1.5604026845637584, + "flow_matching_loss": 0.02408020496368408, + "grad_norm": 0.3091691732406616, + "learning_rate": 2.85126876976474e-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": 0.012264856789261102, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.00596089877653867, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.030490612238645555, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.02408020496368408, + "weighted_rkd_loss": 0.015245306119322777 + }, + { + "epoch": 1.5771812080536913, + "flow_matching_loss": 0.025282399170100688, + "grad_norm": 0.25739333033561707, + "learning_rate": 2.8469322313067726e-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": 0.015094980597496033, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0074165277415886525, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.037606488727033135, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.025282399170100688, + "weighted_rkd_loss": 0.018803244363516568 + }, + { + "epoch": 1.5939597315436242, + "flow_matching_loss": 0.026252748258411885, + "grad_norm": 0.2987407445907593, + "learning_rate": 2.8425367736695522e-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": 0.011627537570893765, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.005372069217264653, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.028627144172787666, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.026252748258411885, + "weighted_rkd_loss": 0.014313572086393833 + }, + { + "epoch": 1.610738255033557, + "flow_matching_loss": 0.02298819851130247, + "grad_norm": 0.2788366973400116, + "learning_rate": 2.838082589124636e-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": 0.014043763000518083, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.006708415551111102, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.034795941412448884, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.02298819851130247, + "weighted_rkd_loss": 0.017397970706224442 + }, + { + "epoch": 1.62751677852349, + "flow_matching_loss": 0.02329830452799797, + "grad_norm": 0.321382999420166, + "learning_rate": 2.8335698725124843e-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": 0.013656491879373789, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.006472870148718357, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.03378585446625948, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.02329830452799797, + "weighted_rkd_loss": 0.01689292723312974 + }, + { + "epoch": 1.6442953020134228, + "flow_matching_loss": 0.022350782155990602, + "grad_norm": 0.2501571476459503, + "learning_rate": 2.8289988212339422e-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": 0.012651610746979713, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.005716795939952135, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.031020017340779305, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.022350782155990602, + "weighted_rkd_loss": 0.015510008670389652 + }, + { + "epoch": 1.6610738255033557, + "flow_matching_loss": 0.021388859674334525, + "grad_norm": 0.21727903187274933, + "learning_rate": 2.8243696352415997e-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": 0.013748683407902718, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.006664992216974497, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.03416235968470573, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.021388859674334525, + "weighted_rkd_loss": 0.017081179842352866 + }, + { + "epoch": 1.6778523489932886, + "flow_matching_loss": 0.025073782540857792, + "grad_norm": 0.2569715678691864, + "learning_rate": 2.8196825170310495e-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": 0.01149385841563344, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.00539817709941417, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.028385894000530244, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.025073782540857792, + "weighted_rkd_loss": 0.014192947000265122 + }, + { + "epoch": 1.6946308724832215, + "flow_matching_loss": 0.025097819790244103, + "grad_norm": 0.27531710267066956, + "learning_rate": 2.8149376716320256e-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": 0.013829244673252106, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.006092684157192707, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.03375117350369692, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.025097819790244103, + "weighted_rkd_loss": 0.01687558675184846 + }, + { + "epoch": 1.7114093959731544, + "flow_matching_loss": 0.022316984459757806, + "grad_norm": 0.27092570066452026, + "learning_rate": 2.8101353065994373e-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": 0.01365465773269534, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.006724758353084326, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.03403407391160727, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.022316984459757806, + "weighted_rkd_loss": 0.017017036955803633 + }, + { + "epoch": 1.7281879194630871, + "flow_matching_loss": 0.02207082649692893, + "grad_norm": 0.2310076355934143, + "learning_rate": 2.8052756320042885e-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": 0.012512798327952624, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.00538945309817791, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.030415049567818642, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.02207082649692893, + "weighted_rkd_loss": 0.015207524783909321 + }, + { + "epoch": 1.7449664429530203, + "flow_matching_loss": 0.02230855505913496, + "grad_norm": 0.25968125462532043, + "learning_rate": 2.8003588604244893e-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": 0.013264590222388507, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.005813045892864466, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.03234222587198019, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.02230855505913496, + "weighted_rkd_loss": 0.016171112935990096 + }, + { + "epoch": 1.761744966442953, + "flow_matching_loss": 0.022760432679206133, + "grad_norm": 0.26551783084869385, + "learning_rate": 2.7953852069355565e-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": 0.01373848868533969, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.00661056051030755, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.034087537229061125, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.022760432679206133, + "weighted_rkd_loss": 0.017043768614530563 + }, + { + "epoch": 1.778523489932886, + "flow_matching_loss": 0.023991832323372365, + "grad_norm": 0.27423810958862305, + "learning_rate": 2.790354889101206e-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": 0.013085504621267318, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.005771422339603305, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.03194243106991053, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.023991832323372365, + "weighted_rkd_loss": 0.015971215534955265 + }, + { + "epoch": 1.7953020134228188, + "flow_matching_loss": 0.024474795162677764, + "grad_norm": 0.27359020709991455, + "learning_rate": 2.7852681269638366e-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": 0.01170850549824536, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.005443004914559424, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.028860015608370305, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.024474795162677764, + "weighted_rkd_loss": 0.014430007804185152 + }, + { + "epoch": 1.8120805369127517, + "flow_matching_loss": 0.021090425737202167, + "grad_norm": 0.22412976622581482, + "learning_rate": 2.780125143034902e-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": 0.01328139128163457, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.006126812007278204, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.03268959484994412, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.021090425737202167, + "weighted_rkd_loss": 0.01634479742497206 + }, + { + "epoch": 1.8288590604026846, + "flow_matching_loss": 0.023801552318036556, + "grad_norm": 0.3204596936702728, + "learning_rate": 2.7749261622851812e-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": 0.012044427543878555, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.006069330545142293, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.03015818577259779, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.023801552318036556, + "weighted_rkd_loss": 0.015079092886298894 + }, + { + "epoch": 1.8456375838926173, + "flow_matching_loss": 0.024525544233620168, + "grad_norm": 0.22625170648097992, + "learning_rate": 2.7696714121349335e-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": 0.011578796617686748, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.005627041519619525, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.028784634731709956, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.024525544233620168, + "weighted_rkd_loss": 0.014392317365854978 + }, + { + "epoch": 1.8624161073825505, + "flow_matching_loss": 0.024563361518085004, + "grad_norm": 0.23123829066753387, + "learning_rate": 2.764361122443954e-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": 0.011116953007876873, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0048352541169151666, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.027069160155951975, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.024563361518085004, + "weighted_rkd_loss": 0.013534580077975988 + }, + { + "epoch": 1.8791946308724832, + "flow_matching_loss": 0.02112237215042114, + "grad_norm": 0.266678124666214, + "learning_rate": 2.758995525501517e-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": 0.010669780941680074, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.005158815323375166, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.02649837750941515, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.02112237215042114, + "weighted_rkd_loss": 0.013249188754707574 + }, + { + "epoch": 1.895973154362416, + "flow_matching_loss": 0.023922207579016686, + "grad_norm": 0.23021627962589264, + "learning_rate": 2.753574856016215e-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": 0.01126117487438023, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.005241665965877473, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.027764015831053257, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.023922207579016686, + "weighted_rkd_loss": 0.013882007915526628 + }, + { + "epoch": 1.912751677852349, + "flow_matching_loss": 0.021286398451775313, + "grad_norm": 0.30017852783203125, + "learning_rate": 2.7480993511056913e-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": 0.012626979872584343, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.00600407593883574, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.03125803600996733, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.021286398451775313, + "weighted_rkd_loss": 0.015629018004983664 + }, + { + "epoch": 1.929530201342282, + "flow_matching_loss": 0.02148480396717787, + "grad_norm": 0.23389174044132233, + "learning_rate": 2.7425692502862684e-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": 0.012474285159260034, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.005805830564349889, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.030754400603473186, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.02148480396717787, + "weighted_rkd_loss": 0.015377200301736593 + }, + { + "epoch": 1.9463087248322148, + "flow_matching_loss": 0.024155576527118684, + "grad_norm": 0.30734431743621826, + "learning_rate": 2.7369847954624728e-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": 0.012808186188340186, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.005811949074268341, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.03142832107841968, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.024155576527118684, + "weighted_rkd_loss": 0.01571416053920984 + }, + { + "epoch": 1.9630872483221475, + "flow_matching_loss": 0.024268321320414545, + "grad_norm": 0.19537605345249176, + "learning_rate": 2.731346230916448e-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": 0.010803611855953932, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.005059144808910787, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.026666368544101714, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.024268321320414545, + "weighted_rkd_loss": 0.013333184272050857 + }, + { + "epoch": 1.9798657718120807, + "flow_matching_loss": 0.021980220358818768, + "grad_norm": 0.29635903239250183, + "learning_rate": 2.7256538032972746e-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": 0.011051633302122355, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.00538994565140456, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.027493212185800076, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.021980220358818768, + "weighted_rkd_loss": 0.013746606092900038 + }, + { + "epoch": 1.9966442953020134, + "flow_matching_loss": 0.01943954061716795, + "grad_norm": 0.2546007037162781, + "learning_rate": 2.7199077616101773e-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": 0.01068065189756453, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.00416237716563046, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.025523680821061136, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.01943954061716795, + "weighted_rkd_loss": 0.012761840410530568 + }, + { + "epoch": 2.0134228187919465, + "flow_matching_loss": 0.021445751748979093, + "grad_norm": 0.2860546410083771, + "learning_rate": 2.714108357205634e-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": 0.012169172149151563, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.005011275131255388, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.029349619150161745, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.021445751748979093, + "weighted_rkd_loss": 0.014674809575080872 + }, + { + "epoch": 2.030201342281879, + "flow_matching_loss": 0.02413941789418459, + "grad_norm": 0.2607485055923462, + "learning_rate": 2.7082558437683803e-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": 0.010609469283372164, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.005363130709156394, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.026582069136202334, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.02413941789418459, + "weighted_rkd_loss": 0.013291034568101167 + }, + { + "epoch": 2.046979865771812, + "flow_matching_loss": 0.02123691262677312, + "grad_norm": 0.2052907645702362, + "learning_rate": 2.702350477306315e-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": 0.011026246706023812, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.004970021010376513, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.027022514678537845, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.02123691262677312, + "weighted_rkd_loss": 0.013511257339268922 + }, + { + "epoch": 2.063758389261745, + "flow_matching_loss": 0.020513943303376435, + "grad_norm": 0.26014190912246704, + "learning_rate": 2.6963925161392963e-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": 0.010161254042759537, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.004656049120239913, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.02497855741530657, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.020513943303376435, + "weighted_rkd_loss": 0.012489278707653284 + }, + { + "epoch": 2.0805369127516777, + "flow_matching_loss": 0.02209649309515953, + "grad_norm": 0.22879625856876373, + "learning_rate": 2.6903822208878476e-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": 0.01278407545760274, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0054352257400751116, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.031003376841545104, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.02209649309515953, + "weighted_rkd_loss": 0.015501688420772552 + }, + { + "epoch": 2.097315436241611, + "flow_matching_loss": 0.020225646905601025, + "grad_norm": 0.19999638199806213, + "learning_rate": 2.6843198544617535e-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": 0.009955055452883244, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.004553360398858786, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.024463471211493015, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.020225646905601025, + "weighted_rkd_loss": 0.012231735605746507 + }, + { + "epoch": 2.1140939597315436, + "flow_matching_loss": 0.025491086300462484, + "grad_norm": 0.3379616141319275, + "learning_rate": 2.6782056820485612e-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": 0.012597405444830656, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.00618452406488359, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.03137933518737555, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.025491086300462484, + "weighted_rkd_loss": 0.015689667593687773 + }, + { + "epoch": 2.1308724832214767, + "flow_matching_loss": 0.022379141207784413, + "grad_norm": 0.24647405743598938, + "learning_rate": 2.672039971101979e-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": 0.01086846049875021, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.00439148754812777, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.026128408685326576, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.022379141207784413, + "weighted_rkd_loss": 0.013064204342663288 + }, + { + "epoch": 2.1476510067114094, + "flow_matching_loss": 0.021409355103969574, + "grad_norm": 0.21077895164489746, + "learning_rate": 2.665822991330178e-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": 0.010168445203453303, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.004441888746805489, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.02477877903729677, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.021409355103969574, + "weighted_rkd_loss": 0.012389389518648385 + }, + { + "epoch": 2.164429530201342, + "flow_matching_loss": 0.022486128844320774, + "grad_norm": 0.2168387919664383, + "learning_rate": 2.6595550146839934e-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": 0.010359025932848454, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0044691324001178145, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.02518718410283327, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.022486128844320774, + "weighted_rkd_loss": 0.012593592051416636 + }, + { + "epoch": 2.1812080536912752, + "flow_matching_loss": 0.02099057734012604, + "grad_norm": 0.2119854986667633, + "learning_rate": 2.6532363153450287e-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": 0.01063207071274519, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.004422961152158678, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.025687102787196638, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.02099057734012604, + "weighted_rkd_loss": 0.012843551393598319 + }, + { + "epoch": 2.197986577181208, + "flow_matching_loss": 0.01913056764751673, + "grad_norm": 0.21082034707069397, + "learning_rate": 2.6468671697136634e-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": 0.009154821233823895, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0036931908456608654, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.022002833429723978, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.01913056764751673, + "weighted_rkd_loss": 0.011001416714861989 + }, + { + "epoch": 2.214765100671141, + "flow_matching_loss": 0.02139302557334304, + "grad_norm": 0.26689302921295166, + "learning_rate": 2.640447856396961e-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": 0.01220207791775465, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.005580418580211699, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.029984574671834706, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.02139302557334304, + "weighted_rkd_loss": 0.014992287335917353 + }, + { + "epoch": 2.2315436241610738, + "flow_matching_loss": 0.02142320163547993, + "grad_norm": 0.19641175866127014, + "learning_rate": 2.6339786561964805e-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": 0.010372415697202087, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.005048008938319981, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.025792840495705606, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.02142320163547993, + "weighted_rkd_loss": 0.012896420247852803 + }, + { + "epoch": 2.248322147651007, + "flow_matching_loss": 0.019434516970068216, + "grad_norm": 0.2214556634426117, + "learning_rate": 2.6274598520959966e-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": 0.01033841255120933, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.004615865182131529, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.02529268991202116, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.019434516970068216, + "weighted_rkd_loss": 0.01264634495601058 + }, + { + "epoch": 2.2651006711409396, + "flow_matching_loss": 0.02066937694326043, + "grad_norm": 0.19580107927322388, + "learning_rate": 2.6208917292491185e-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": 0.010874006384983658, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.004946777364239097, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.02669478990137577, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.02066937694326043, + "weighted_rkd_loss": 0.013347394950687885 + }, + { + "epoch": 2.2818791946308723, + "flow_matching_loss": 0.02022442305460572, + "grad_norm": 0.19812536239624023, + "learning_rate": 2.6142745749668176e-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": 0.011259743850678206, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.005238024378195405, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.027757512405514717, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.02022442305460572, + "weighted_rkd_loss": 0.013878756202757359 + }, + { + "epoch": 2.2986577181208054, + "flow_matching_loss": 0.020008648745715618, + "grad_norm": 0.3320487439632416, + "learning_rate": 2.607608678704859e-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": 0.011717541143298149, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.005797909549437463, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.02923299167305231, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.020008648745715618, + "weighted_rkd_loss": 0.014616495836526156 + }, + { + "epoch": 2.315436241610738, + "flow_matching_loss": 0.019694368913769723, + "grad_norm": 0.2613303065299988, + "learning_rate": 2.600894332051139e-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": 0.009247061843052507, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.004215921042487025, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.022710044495761396, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.019694368913769723, + "weighted_rkd_loss": 0.011355022247880698 + }, + { + "epoch": 2.3322147651006713, + "flow_matching_loss": 0.023736210353672505, + "grad_norm": 0.20574374496936798, + "learning_rate": 2.594131828712932e-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": 0.010481608007103205, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.004885053401812911, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.025848269648849963, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.023736210353672505, + "weighted_rkd_loss": 0.012924134824424982 + }, + { + "epoch": 2.348993288590604, + "flow_matching_loss": 0.020662209298461676, + "grad_norm": 0.23922887444496155, + "learning_rate": 2.5873214645040414e-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": 0.010238038096576928, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.004725828301161528, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.025201904587447643, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.020662209298461676, + "weighted_rkd_loss": 0.012600952293723822 + }, + { + "epoch": 2.365771812080537, + "flow_matching_loss": 0.020602713525295257, + "grad_norm": 0.3683367967605591, + "learning_rate": 2.5804635373318604e-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": 0.009888650011271238, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.004859140398912132, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.024636439979076385, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.020602713525295257, + "weighted_rkd_loss": 0.012318219989538193 + }, + { + "epoch": 2.38255033557047, + "flow_matching_loss": 0.021966058015823364, + "grad_norm": 0.2076294869184494, + "learning_rate": 2.57355834718434e-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": 0.01116554499603808, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.004525258555077016, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.026856348384171724, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.021966058015823364, + "weighted_rkd_loss": 0.013428174192085862 + }, + { + "epoch": 2.3993288590604025, + "flow_matching_loss": 0.019420094788074493, + "grad_norm": 0.22278212010860443, + "learning_rate": 2.566606196116867e-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": 0.009492038935422897, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.004202829161658883, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.02318690689280629, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.019420094788074493, + "weighted_rkd_loss": 0.011593453446403145 + }, + { + "epoch": 2.4161073825503356, + "flow_matching_loss": 0.02243018485605717, + "grad_norm": 0.23856237530708313, + "learning_rate": 2.559607388239051e-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": 0.011394480243325233, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.005045613972470164, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.027834573946893214, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.02243018485605717, + "weighted_rkd_loss": 0.013917286973446607 + }, + { + "epoch": 2.4328859060402683, + "flow_matching_loss": 0.018066379148513077, + "grad_norm": 0.24139682948589325, + "learning_rate": 2.552562229701422e-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": 0.010150557290762663, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.004355087480507791, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.02465620171278715, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.018066379148513077, + "weighted_rkd_loss": 0.012328100856393576 + }, + { + "epoch": 2.4496644295302015, + "flow_matching_loss": 0.020288294926285744, + "grad_norm": 0.20675493776798248, + "learning_rate": 2.545471028682036e-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": 0.009677186142653227, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.004701676638796925, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.02405604887753725, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.020288294926285744, + "weighted_rkd_loss": 0.012028024438768626 + }, + { + "epoch": 2.466442953020134, + "flow_matching_loss": 0.022038421034812926, + "grad_norm": 0.25239408016204834, + "learning_rate": 2.538334095373e-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": 0.009866706281900405, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.004709447175264359, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.024442859552800657, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.022038421034812926, + "weighted_rkd_loss": 0.012221429776400328 + }, + { + "epoch": 2.4832214765100673, + "flow_matching_loss": 0.02074492024257779, + "grad_norm": 0.2533192038536072, + "learning_rate": 2.531151741966896e-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": 0.010699263401329517, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.005152890039607882, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.026551416702568532, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.02074492024257779, + "weighted_rkd_loss": 0.013275708351284266 + }, + { + "epoch": 2.5, + "flow_matching_loss": 0.021557819843292237, + "grad_norm": 0.2374899834394455, + "learning_rate": 2.5239242826431302e-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": 0.009755126433447003, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0043184555834159255, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.02382870838046074, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.021557819843292237, + "weighted_rkd_loss": 0.01191435419023037 + }, + { + "epoch": 2.5167785234899327, + "flow_matching_loss": 0.021232588775455953, + "grad_norm": 0.20108439028263092, + "learning_rate": 2.5166520335541865e-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": 0.009910681471228599, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.004524847003631293, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.0243462098762393, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.021232588775455953, + "weighted_rkd_loss": 0.01217310493811965 + }, + { + "epoch": 2.533557046979866, + "flow_matching_loss": 0.018248109705746175, + "grad_norm": 0.27282431721687317, + "learning_rate": 2.5093353128117996e-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": 0.010254707466810942, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.004656481766141951, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.02516589667648077, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.018248109705746175, + "weighted_rkd_loss": 0.012582948338240385 + }, + { + "epoch": 2.5503355704697985, + "flow_matching_loss": 0.018883543834090234, + "grad_norm": 0.17541445791721344, + "learning_rate": 2.501974440473037e-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": 0.009280018880963326, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.004575668601319194, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.023135706316679716, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.018883543834090234, + "weighted_rkd_loss": 0.011567853158339858 + }, + { + "epoch": 2.5671140939597317, + "flow_matching_loss": 0.0226139135658741, + "grad_norm": 0.23777875304222107, + "learning_rate": 2.4945697385263015e-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": 0.009309209184721112, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.004227712587453425, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.022846130840480327, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.0226139135658741, + "weighted_rkd_loss": 0.011423065420240163 + }, + { + "epoch": 2.5838926174496644, + "flow_matching_loss": 0.0199148072861135, + "grad_norm": 0.2086194008588791, + "learning_rate": 2.487121530877243e-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": 0.010670735454186796, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.004725530091673136, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.026067000441253187, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.0199148072861135, + "weighted_rkd_loss": 0.013033500220626593 + }, + { + "epoch": 2.600671140939597, + "flow_matching_loss": 0.01935564409941435, + "grad_norm": 0.24306678771972656, + "learning_rate": 2.4796301433345943e-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": 0.009589324938133359, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.004376536863856017, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.0235551867634058, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.01935564409941435, + "weighted_rkd_loss": 0.0117775933817029 + }, + { + "epoch": 2.61744966442953, + "flow_matching_loss": 0.019802154507488012, + "grad_norm": 0.218539297580719, + "learning_rate": 2.4720959035959148e-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": 0.010907125566154719, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.004833343508653343, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.026647594571113587, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.019802154507488012, + "weighted_rkd_loss": 0.013323797285556794 + }, + { + "epoch": 2.6342281879194633, + "flow_matching_loss": 0.020271860156208278, + "grad_norm": 0.2485092729330063, + "learning_rate": 2.464519141233257e-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": 0.010241875238716603, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0045910695334896445, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.025074820034205914, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.020271860156208278, + "weighted_rkd_loss": 0.012537410017102957 + }, + { + "epoch": 2.651006711409396, + "flow_matching_loss": 0.018997634947299957, + "grad_norm": 0.2614387571811676, + "learning_rate": 2.456900187678754e-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": 0.00964788207784295, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.004860389954410493, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.024156153947114945, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.018997634947299957, + "weighted_rkd_loss": 0.012078076973557473 + }, + { + "epoch": 2.6677852348993287, + "flow_matching_loss": 0.01944811325520277, + "grad_norm": 0.19716313481330872, + "learning_rate": 2.4492393762101147e-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": 0.0073358194902539255, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0034736467525362967, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.018145285733044147, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.01944811325520277, + "weighted_rkd_loss": 0.009072642866522073 + }, + { + "epoch": 2.684563758389262, + "flow_matching_loss": 0.020127653423696757, + "grad_norm": 0.23337966203689575, + "learning_rate": 2.4415370419360508e-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": 0.008881192840635776, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0041180340806022285, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.021880419738590718, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.020127653423696757, + "weighted_rkd_loss": 0.010940209869295359 + }, + { + "epoch": 2.7013422818791946, + "flow_matching_loss": 0.019465051591396332, + "grad_norm": 0.2043837606906891, + "learning_rate": 2.4337935217816157e-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": 0.00927969217300415, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.004235364822670817, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.022794749494642018, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.019465051591396332, + "weighted_rkd_loss": 0.011397374747321009 + }, + { + "epoch": 2.7181208053691277, + "flow_matching_loss": 0.022998061403632163, + "grad_norm": 0.3106498718261719, + "learning_rate": 2.4260091544734665e-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": 0.00990912769921124, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.004301568167284131, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.024119823798537256, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.022998061403632163, + "weighted_rkd_loss": 0.012059911899268628 + }, + { + "epoch": 2.7348993288590604, + "flow_matching_loss": 0.021626121085137128, + "grad_norm": 0.2143699675798416, + "learning_rate": 2.4181842805250466e-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": 0.008956945640966297, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.004163260711356998, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.02207715194672346, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.021626121085137128, + "weighted_rkd_loss": 0.01103857597336173 + }, + { + "epoch": 2.751677852348993, + "flow_matching_loss": 0.018923975061625242, + "grad_norm": 0.2210591435432434, + "learning_rate": 2.4103192422216923e-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": 0.008583794254809618, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.003714717063121498, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.02088230550289154, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.018923975061625242, + "weighted_rkd_loss": 0.01044115275144577 + }, + { + "epoch": 2.7684563758389262, + "flow_matching_loss": 0.019391734432429076, + "grad_norm": 0.24085281789302826, + "learning_rate": 2.402414383605658e-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": 0.00893097072839737, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0037688536569476126, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.021630795020610095, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.019391734432429076, + "weighted_rkd_loss": 0.010815397510305047 + }, + { + "epoch": 2.785234899328859, + "flow_matching_loss": 0.021253248024731875, + "grad_norm": 0.21485064923763275, + "learning_rate": 2.3944700504610674e-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": 0.00990743380971253, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.003974048513919115, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.0237889158539474, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.021253248024731875, + "weighted_rkd_loss": 0.0118944579269737 + }, + { + "epoch": 2.802013422818792, + "flow_matching_loss": 0.01799254035577178, + "grad_norm": 0.21391142904758453, + "learning_rate": 2.3864865902987883e-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": 0.007719148229807615, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0034132912289351223, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01885158782824874, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.01799254035577178, + "weighted_rkd_loss": 0.00942579391412437 + }, + { + "epoch": 2.8187919463087248, + "flow_matching_loss": 0.02177695669233799, + "grad_norm": 0.1809767782688141, + "learning_rate": 2.378464352341231e-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": 0.007593847950920463, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.003742655669339001, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01893035164102912, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.02177695669233799, + "weighted_rkd_loss": 0.00946517582051456 + }, + { + "epoch": 2.8355704697986575, + "flow_matching_loss": 0.018650972004979848, + "grad_norm": 0.15258850157260895, + "learning_rate": 2.3704036875070728e-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": 0.008111832011491061, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.003444994508754462, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.0196686583571136, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.018650972004979848, + "weighted_rkd_loss": 0.0098343291785568 + }, + { + "epoch": 2.8523489932885906, + "flow_matching_loss": 0.021410140488296748, + "grad_norm": 0.2503313720226288, + "learning_rate": 2.362304948395906e-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": 0.008499804651364684, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.004048812622204423, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.021048422157764434, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.021410140488296748, + "weighted_rkd_loss": 0.010524211078882217 + }, + { + "epoch": 2.8691275167785237, + "flow_matching_loss": 0.018689179420471193, + "grad_norm": 0.2221410572528839, + "learning_rate": 2.3541684892728157e-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": 0.00805930569767952, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0035608888836577536, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01967950016260147, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.018689179420471193, + "weighted_rkd_loss": 0.009839750081300735 + }, + { + "epoch": 2.8859060402684564, + "flow_matching_loss": 0.021520638652145863, + "grad_norm": 0.22633783519268036, + "learning_rate": 2.345994666052883e-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": 0.009966268250718714, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.004593332833610475, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.024525869358330964, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.021520638652145863, + "weighted_rkd_loss": 0.012262934679165482 + }, + { + "epoch": 2.902684563758389, + "flow_matching_loss": 0.019396231137216093, + "grad_norm": 0.17181359231472015, + "learning_rate": 2.337783836285614e-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": 0.007831518398597836, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.00329853689763695, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.018961573857814072, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.019396231137216093, + "weighted_rkd_loss": 0.009480786928907036 + }, + { + "epoch": 2.9194630872483223, + "flow_matching_loss": 0.01790900407359004, + "grad_norm": 0.22203585505485535, + "learning_rate": 2.3295363591393026e-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": 0.009272311627864838, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0037968565709888937, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.022341480012983084, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.01790900407359004, + "weighted_rkd_loss": 0.011170740006491542 + }, + { + "epoch": 2.936241610738255, + "flow_matching_loss": 0.020569243282079697, + "grad_norm": 0.25884026288986206, + "learning_rate": 2.3212525953853174e-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": 0.008792810002341866, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.004242602386511862, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.021828222274780273, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.020569243282079697, + "weighted_rkd_loss": 0.010914111137390136 + }, + { + "epoch": 2.953020134228188, + "flow_matching_loss": 0.019374095369130373, + "grad_norm": 0.2063402235507965, + "learning_rate": 2.3129329073823206e-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": 0.007920988649129868, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0032342170015908778, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.019076194148510695, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.019374095369130373, + "weighted_rkd_loss": 0.009538097074255347 + }, + { + "epoch": 2.969798657718121, + "flow_matching_loss": 0.020364384446293116, + "grad_norm": 0.24272602796554565, + "learning_rate": 2.3045776590604175e-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": 0.008952021831646562, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.004419904644601047, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.02232394814491272, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.020364384446293116, + "weighted_rkd_loss": 0.01116197407245636 + }, + { + "epoch": 2.9865771812080535, + "flow_matching_loss": 0.019254895765334366, + "grad_norm": 0.24223721027374268, + "learning_rate": 2.296187215905237e-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": 0.008203888079151511, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.003970045782625675, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.020377822034060956, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.019254895765334366, + "weighted_rkd_loss": 0.010188911017030478 + }, + { + "epoch": 3.0033557046979866, + "flow_matching_loss": 0.020528820529580116, + "grad_norm": 0.20469991862773895, + "learning_rate": 2.2877619449419437e-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": 0.007477659825235605, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.003407218540087342, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.018362538050860168, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.020528820529580116, + "weighted_rkd_loss": 0.009181269025430084 + }, + { + "epoch": 3.0201342281879193, + "flow_matching_loss": 0.019945137668401003, + "grad_norm": 0.2324223518371582, + "learning_rate": 2.2793022147191843e-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": 0.007709356397390366, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.003613951010629535, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.019032663851976394, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.019945137668401003, + "weighted_rkd_loss": 0.009516331925988197 + }, + { + "epoch": 3.0369127516778525, + "flow_matching_loss": 0.019441381469368933, + "grad_norm": 0.20990999042987823, + "learning_rate": 2.2708083952929634e-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": 0.008489918569102884, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.003641888406127691, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.020621725637465716, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.019441381469368933, + "weighted_rkd_loss": 0.010310862818732858 + }, + { + "epoch": 3.053691275167785, + "flow_matching_loss": 0.020685941353440283, + "grad_norm": 0.20463426411151886, + "learning_rate": 2.2622808582104594e-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": 0.008092707209289073, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0032459887210279704, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01943140309303999, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.020685941353440283, + "weighted_rkd_loss": 0.009715701546519994 + }, + { + "epoch": 3.0704697986577183, + "flow_matching_loss": 0.017734704166650773, + "grad_norm": 0.19989708065986633, + "learning_rate": 2.2537199764937703e-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": 0.008762143971398472, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.003955869167111814, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.021480157133191825, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.017734704166650773, + "weighted_rkd_loss": 0.010740078566595913 + }, + { + "epoch": 3.087248322147651, + "flow_matching_loss": 0.019546301569789648, + "grad_norm": 0.18527752161026, + "learning_rate": 2.245126124623595e-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": 0.007820900995284319, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0036649121902883053, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.019306713808327915, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.019546301569789648, + "weighted_rkd_loss": 0.009653356904163957 + }, + { + "epoch": 3.1040268456375837, + "flow_matching_loss": 0.019912993628531696, + "grad_norm": 0.20761150121688843, + "learning_rate": 2.2364996785228547e-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": 0.008325929567217828, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0037194554693996905, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.020371314603835344, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.019912993628531696, + "weighted_rkd_loss": 0.010185657301917672 + }, + { + "epoch": 3.120805369127517, + "flow_matching_loss": 0.01854344429448247, + "grad_norm": 0.27597880363464355, + "learning_rate": 2.227841015540248e-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": 0.007440061867237091, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0033744079642929136, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.018254531640559436, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.01854344429448247, + "weighted_rkd_loss": 0.009127265820279718 + }, + { + "epoch": 3.1375838926174495, + "flow_matching_loss": 0.01822344996035099, + "grad_norm": 0.2510194182395935, + "learning_rate": 2.2191505144337433e-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": 0.008060201490297913, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0034365889034233986, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01955699184909463, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.01822344996035099, + "weighted_rkd_loss": 0.009778495924547315 + }, + { + "epoch": 3.1543624161073827, + "flow_matching_loss": 0.018472170550376178, + "grad_norm": 0.2085818201303482, + "learning_rate": 2.210428555354013e-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": 0.00899176225066185, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.004113292112015188, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.022096816543489696, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.018472170550376178, + "weighted_rkd_loss": 0.011048408271744848 + }, + { + "epoch": 3.1711409395973154, + "flow_matching_loss": 0.01865291325375438, + "grad_norm": 0.24403251707553864, + "learning_rate": 2.201675519827802e-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": 0.00967541616410017, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.004230940202251077, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.023581773135811092, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.01865291325375438, + "weighted_rkd_loss": 0.011790886567905546 + }, + { + "epoch": 3.1879194630872485, + "flow_matching_loss": 0.019162445422261955, + "grad_norm": 0.19958388805389404, + "learning_rate": 2.1928917907412403e-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": 0.00812592231668532, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.003293149534147233, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.019544994179159402, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.019162445422261955, + "weighted_rkd_loss": 0.009772497089579701 + }, + { + "epoch": 3.204697986577181, + "flow_matching_loss": 0.020222001243382692, + "grad_norm": 0.22923623025417328, + "learning_rate": 2.1840777523230937e-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": 0.008355605648830533, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.00404962741304189, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.02076083840802312, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.020222001243382692, + "weighted_rkd_loss": 0.01038041920401156 + }, + { + "epoch": 3.221476510067114, + "flow_matching_loss": 0.017797662504017354, + "grad_norm": 0.16626043617725372, + "learning_rate": 2.1752337901279552e-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": 0.007557774567976594, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0030893665505573155, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.018204915709793566, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.017797662504017354, + "weighted_rkd_loss": 0.009102457854896783 + }, + { + "epoch": 3.238255033557047, + "flow_matching_loss": 0.018512923270463943, + "grad_norm": 0.27996185421943665, + "learning_rate": 2.166360291019383e-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": 0.009122877335175871, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.003923690970987081, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.02216944545507431, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.018512923270463943, + "weighted_rkd_loss": 0.011084722727537155 + }, + { + "epoch": 3.2550335570469797, + "flow_matching_loss": 0.017441819049417974, + "grad_norm": 0.18821027874946594, + "learning_rate": 2.157457643152973e-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": 0.007844586251303553, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.003382657514885068, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.019071830064058305, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.017441819049417974, + "weighted_rkd_loss": 0.009535915032029153 + }, + { + "epoch": 3.271812080536913, + "flow_matching_loss": 0.020743667520582675, + "grad_norm": 0.17960694432258606, + "learning_rate": 2.148526235959385e-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": 0.007514700060710311, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.003746157558634877, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.018775557819753885, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.020743667520582675, + "weighted_rkd_loss": 0.009387778909876942 + }, + { + "epoch": 3.2885906040268456, + "flow_matching_loss": 0.01746414937078953, + "grad_norm": 0.21814922988414764, + "learning_rate": 2.1395664601273016e-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": 0.00813716514967382, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.003756172489374876, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.020030502881854772, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.01746414937078953, + "weighted_rkd_loss": 0.010015251440927386 + }, + { + "epoch": 3.3053691275167787, + "flow_matching_loss": 0.01928861755877733, + "grad_norm": 0.20099057257175446, + "learning_rate": 2.130578707586345e-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": 0.009006292605772615, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.004282481945119798, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.02229506690055132, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.01928861755877733, + "weighted_rkd_loss": 0.01114753345027566 + }, + { + "epoch": 3.3221476510067114, + "flow_matching_loss": 0.019633220881223677, + "grad_norm": 0.32565194368362427, + "learning_rate": 2.1215633714899264e-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": 0.007109207799658179, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.003497364791110158, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01771578062325716, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.019633220881223677, + "weighted_rkd_loss": 0.00885789031162858 + }, + { + "epoch": 3.3389261744966445, + "flow_matching_loss": 0.019557264074683188, + "grad_norm": 0.23021483421325684, + "learning_rate": 2.112520846198053e-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": 0.00938012171536684, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.004018020024523139, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.022778263688087462, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.019557264074683188, + "weighted_rkd_loss": 0.011389131844043731 + }, + { + "epoch": 3.3557046979865772, + "flow_matching_loss": 0.018558720592409372, + "grad_norm": 0.23532751202583313, + "learning_rate": 2.103451527260074e-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": 0.0072332804556936026, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.003332280437462032, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.0177988413721323, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.018558720592409372, + "weighted_rkd_loss": 0.00889942068606615 + }, + { + "epoch": 3.37248322147651, + "flow_matching_loss": 0.016960114892572164, + "grad_norm": 0.1779426634311676, + "learning_rate": 2.094355811397381e-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": 0.00720683760009706, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0033162130159325896, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.017729888390749692, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.016960114892572164, + "weighted_rkd_loss": 0.008864944195374846 + }, + { + "epoch": 3.389261744966443, + "flow_matching_loss": 0.0223352144472301, + "grad_norm": 0.18216963112354279, + "learning_rate": 2.0852340964860522e-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": 0.007228753669187427, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.00323237837292254, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01768988585099578, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.0223352144472301, + "weighted_rkd_loss": 0.00884494292549789 + }, + { + "epoch": 3.4060402684563758, + "flow_matching_loss": 0.01868910351768136, + "grad_norm": 0.25047680735588074, + "learning_rate": 2.076086781539447e-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": 0.00634731762111187, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.002792268805205822, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.015486904233694077, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.01868910351768136, + "weighted_rkd_loss": 0.007743452116847039 + }, + { + "epoch": 3.422818791946309, + "flow_matching_loss": 0.01628869567066431, + "grad_norm": 0.21377144753932953, + "learning_rate": 2.066914266690755e-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": 0.0076282296795398, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0031974301673471928, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.018453889526426792, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.01628869567066431, + "weighted_rkd_loss": 0.009226944763213396 + }, + { + "epoch": 3.4395973154362416, + "flow_matching_loss": 0.017802555300295352, + "grad_norm": 0.2119053453207016, + "learning_rate": 2.0577169531754906e-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": 0.008745979797095061, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0036584984976798297, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.021150457952171565, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.017802555300295352, + "weighted_rkd_loss": 0.010575228976085782 + }, + { + "epoch": 3.4563758389261743, + "flow_matching_loss": 0.02128584710881114, + "grad_norm": 0.22231817245483398, + "learning_rate": 2.048495243313942e-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": 0.006234188843518495, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.002860723249614239, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01532910093665123, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.02128584710881114, + "weighted_rkd_loss": 0.007664550468325615 + }, + { + "epoch": 3.4731543624161074, + "flow_matching_loss": 0.01747396485880017, + "grad_norm": 0.16283448040485382, + "learning_rate": 2.039249540493572e-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": 0.008004908030852676, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.003741130093112588, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.019750946201384068, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.01747396485880017, + "weighted_rkd_loss": 0.009875473100692034 + }, + { + "epoch": 3.48993288590604, + "flow_matching_loss": 0.016751274187117814, + "grad_norm": 0.16140390932559967, + "learning_rate": 2.0299802491513755e-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": 0.008392594195902348, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0036155674140900375, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.020400756038725376, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.016751274187117814, + "weighted_rkd_loss": 0.010200378019362688 + }, + { + "epoch": 3.5067114093959733, + "flow_matching_loss": 0.0161241271533072, + "grad_norm": 0.19097782671451569, + "learning_rate": 2.0206877747561828e-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": 0.00715286280028522, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.003387082903645933, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.017692808248102664, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.0161241271533072, + "weighted_rkd_loss": 0.008846404124051332 + }, + { + "epoch": 3.523489932885906, + "flow_matching_loss": 0.01573560368269682, + "grad_norm": 0.15962116420269012, + "learning_rate": 2.011372523790927e-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": 0.007532580941915512, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0030532466829754413, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.018118408601731063, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.01573560368269682, + "weighted_rkd_loss": 0.009059204300865532 + }, + { + "epoch": 3.540268456375839, + "flow_matching_loss": 0.017809078842401505, + "grad_norm": 0.21319308876991272, + "learning_rate": 2.0020349037348643e-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": 0.006576060969382525, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.002980366419069469, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.016132488194853068, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.017809078842401505, + "weighted_rkd_loss": 0.008066244097426534 + }, + { + "epoch": 3.557046979865772, + "flow_matching_loss": 0.016354834008961917, + "grad_norm": 0.2578032612800598, + "learning_rate": 1.9926753230457443e-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": 0.007642744854092598, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.003393427538685501, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.018678917177021505, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.016354834008961917, + "weighted_rkd_loss": 0.009339458588510752 + }, + { + "epoch": 3.573825503355705, + "flow_matching_loss": 0.021757231652736665, + "grad_norm": 0.20611248910427094, + "learning_rate": 1.983294191141948e-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": 0.008212018711492419, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0035262683406472204, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.019950305856764317, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.021757231652736665, + "weighted_rkd_loss": 0.009975152928382159 + }, + { + "epoch": 3.5906040268456376, + "flow_matching_loss": 0.018257381860166788, + "grad_norm": 0.18068057298660278, + "learning_rate": 1.973891918384575e-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": 0.0075719773303717375, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.002988309320062399, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.018132264073938132, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.018257381860166788, + "weighted_rkd_loss": 0.009066132036969066 + }, + { + "epoch": 3.6073825503355703, + "flow_matching_loss": 0.02030982803553343, + "grad_norm": 0.19641844928264618, + "learning_rate": 1.9644689160594945e-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": 0.007672074157744646, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0032949109561741354, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.018639058992266654, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.02030982803553343, + "weighted_rkd_loss": 0.009319529496133327 + }, + { + "epoch": 3.6241610738255035, + "flow_matching_loss": 0.01583105819299817, + "grad_norm": 0.196472629904747, + "learning_rate": 1.9550255963593553e-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": 0.00726749338209629, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0029815741581842303, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.017516561038792133, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.01583105819299817, + "weighted_rkd_loss": 0.008758280519396066 + }, + { + "epoch": 3.640939597315436, + "flow_matching_loss": 0.017667799256742, + "grad_norm": 0.22149547934532166, + "learning_rate": 1.9455623723655524e-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": 0.007614389341324568, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0031435619108378886, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.018372340593487024, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.017667799256742, + "weighted_rkd_loss": 0.009186170296743512 + }, + { + "epoch": 3.6577181208053693, + "flow_matching_loss": 0.01834620479494333, + "grad_norm": 0.20028650760650635, + "learning_rate": 1.9360796580301606e-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": 0.007079559192061424, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.00302410339936614, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.017183221876621246, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.01834620479494333, + "weighted_rkd_loss": 0.008591610938310623 + }, + { + "epoch": 3.674496644295302, + "flow_matching_loss": 0.020137379597872495, + "grad_norm": 0.17236366868019104, + "learning_rate": 1.9265778681578245e-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": 0.007511051744222641, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.003298318339511752, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.018320421874523162, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.020137379597872495, + "weighted_rkd_loss": 0.009160210937261581 + }, + { + "epoch": 3.6912751677852347, + "flow_matching_loss": 0.018215846829116346, + "grad_norm": 0.2469547539949417, + "learning_rate": 1.9170574183876143e-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": 0.006901636626571417, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0030099719297140837, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.016813245229423047, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.018215846829116346, + "weighted_rkd_loss": 0.008406622614711524 + }, + { + "epoch": 3.708053691275168, + "flow_matching_loss": 0.015153492335230112, + "grad_norm": 0.14944574236869812, + "learning_rate": 1.9075187251748452e-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": 0.006793143972754478, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.003026555338874459, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01661284323781729, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.015153492335230112, + "weighted_rkd_loss": 0.008306421618908644 + }, + { + "epoch": 3.7248322147651005, + "flow_matching_loss": 0.017565101385116577, + "grad_norm": 0.15106534957885742, + "learning_rate": 1.8979622057728596e-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": 0.007395245274528861, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.003232283913530409, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.018022774532437325, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.017565101385116577, + "weighted_rkd_loss": 0.009011387266218663 + }, + { + "epoch": 3.7416107382550337, + "flow_matching_loss": 0.015845688339322806, + "grad_norm": 0.17569081485271454, + "learning_rate": 1.8883882782147748e-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": 0.007898155646398664, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0035057956352829932, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01930210692808032, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.015845688339322806, + "weighted_rkd_loss": 0.00965105346404016 + }, + { + "epoch": 3.7583892617449663, + "flow_matching_loss": 0.017872312851250172, + "grad_norm": 0.19289667904376984, + "learning_rate": 1.8787973612951967e-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": 0.00726058604195714, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0030537315062247215, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01757490374147892, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.017872312851250172, + "weighted_rkd_loss": 0.00878745187073946 + }, + { + "epoch": 3.7751677852348995, + "flow_matching_loss": 0.017769175115972757, + "grad_norm": 0.21152235567569733, + "learning_rate": 1.8691898745519028e-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": 0.0071007994469255205, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0031611404730938373, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.017362739052623512, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.017769175115972757, + "weighted_rkd_loss": 0.008681369526311756 + }, + { + "epoch": 3.791946308724832, + "flow_matching_loss": 0.019934184104204177, + "grad_norm": 0.18968239426612854, + "learning_rate": 1.8595662382474863e-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": 0.006336328200995922, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.002998991427011788, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.015671647805720567, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.019934184104204177, + "weighted_rkd_loss": 0.007835823902860284 + }, + { + "epoch": 3.8087248322147653, + "flow_matching_loss": 0.015440336987376213, + "grad_norm": 0.20246362686157227, + "learning_rate": 1.8499268733509766e-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": 0.006958040781319142, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.003026825119741261, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01694290665909648, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.015440336987376213, + "weighted_rkd_loss": 0.00847145332954824 + }, + { + "epoch": 3.825503355704698, + "flow_matching_loss": 0.019549896754324437, + "grad_norm": 0.20109418034553528, + "learning_rate": 1.84027220151942e-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": 0.007252618251368403, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0030437311856076122, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01754896780475974, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.019549896754324437, + "weighted_rkd_loss": 0.00877448390237987 + }, + { + "epoch": 3.8422818791946307, + "flow_matching_loss": 0.017891636677086355, + "grad_norm": 0.2103230506181717, + "learning_rate": 1.83060264507944e-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": 0.006949930777773261, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.002904061577282846, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.016803923156112434, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.017891636677086355, + "weighted_rkd_loss": 0.008401961578056217 + }, + { + "epoch": 3.859060402684564, + "flow_matching_loss": 0.015393044054508209, + "grad_norm": 0.17709612846374512, + "learning_rate": 1.82091862700876e-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": 0.008440901571884752, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0034887955291196702, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.020370598416775465, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.015393044054508209, + "weighted_rkd_loss": 0.010185299208387733 + }, + { + "epoch": 3.8758389261744965, + "flow_matching_loss": 0.01742718815803528, + "grad_norm": 0.192051962018013, + "learning_rate": 1.8112205709177023e-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": 0.007235960615798831, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0028274517389945686, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.017299372982233763, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.01742718815803528, + "weighted_rkd_loss": 0.008649686491116882 + }, + { + "epoch": 3.8926174496644297, + "flow_matching_loss": 0.016846919432282448, + "grad_norm": 0.21569876372814178, + "learning_rate": 1.8015089010306585e-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": 0.006895868247374893, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.002992071001790464, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.016783807333558796, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.016846919432282448, + "weighted_rkd_loss": 0.008391903666779398 + }, + { + "epoch": 3.9093959731543624, + "flow_matching_loss": 0.016210787929594517, + "grad_norm": 0.15277397632598877, + "learning_rate": 1.7917840421675298e-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": 0.0056195994140580295, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.002441662212368101, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.013680860958993435, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.016210787929594517, + "weighted_rkd_loss": 0.006840430479496717 + }, + { + "epoch": 3.926174496644295, + "flow_matching_loss": 0.015086184721440077, + "grad_norm": 0.17707772552967072, + "learning_rate": 1.7820464197251474e-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": 0.0054072860395535825, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.002303721779026091, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.013118293881416321, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.015086184721440077, + "weighted_rkd_loss": 0.006559146940708161 + }, + { + "epoch": 3.942953020134228, + "flow_matching_loss": 0.020812822692096233, + "grad_norm": 0.22611883282661438, + "learning_rate": 1.7722964596586627e-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": 0.007290642336010933, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0030042332131415606, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01758551774546504, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.020812822692096233, + "weighted_rkd_loss": 0.00879275887273252 + }, + { + "epoch": 3.959731543624161, + "flow_matching_loss": 0.01684179836884141, + "grad_norm": 0.22449995577335358, + "learning_rate": 1.7625345884629144e-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": 0.006925484677776694, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.003144025569781661, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.016994994692504407, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.01684179836884141, + "weighted_rkd_loss": 0.008497497346252204 + }, + { + "epoch": 3.976510067114094, + "flow_matching_loss": 0.017648623324930668, + "grad_norm": 0.20463936030864716, + "learning_rate": 1.7527612331537722e-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": 0.006672896863892675, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0029675276484340428, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.016313321236521006, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.017648623324930668, + "weighted_rkd_loss": 0.008156660618260503 + }, + { + "epoch": 3.9932885906040267, + "flow_matching_loss": 0.017300537042319774, + "grad_norm": 0.21151456236839294, + "learning_rate": 1.7429768212494593e-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": 0.007149667013436556, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.003181573876645416, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.017480908054858448, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.017300537042319774, + "weighted_rkd_loss": 0.008740454027429224 + }, + { + "epoch": 4.010067114093959, + "flow_matching_loss": 0.015868636220693587, + "grad_norm": 0.1958484798669815, + "learning_rate": 1.7331817807518497e-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": 0.0064332588808611035, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0031656825682148336, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.016032200399786235, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.015868636220693587, + "weighted_rkd_loss": 0.008016100199893118 + }, + { + "epoch": 4.026845637583893, + "flow_matching_loss": 0.01652771672233939, + "grad_norm": 0.2061646282672882, + "learning_rate": 1.723376540127745e-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": 0.007571346079930663, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.003144329390488565, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01828702138736844, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.01652771672233939, + "weighted_rkd_loss": 0.00914351069368422 + }, + { + "epoch": 4.043624161073826, + "flow_matching_loss": 0.01697065941989422, + "grad_norm": 0.13876841962337494, + "learning_rate": 1.7135615282901357e-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": 0.007565911905840039, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0037172524258494376, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.018849076144397257, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.01697065941989422, + "weighted_rkd_loss": 0.009424538072198629 + }, + { + "epoch": 4.060402684563758, + "flow_matching_loss": 0.019290306326001882, + "grad_norm": 0.23693525791168213, + "learning_rate": 1.7037371745794358e-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": 0.006493809865787625, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.002745014219544828, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.015732633974403144, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.019290306326001882, + "weighted_rkd_loss": 0.007866316987201572 + }, + { + "epoch": 4.077181208053691, + "flow_matching_loss": 0.01645718878135085, + "grad_norm": 0.1961735188961029, + "learning_rate": 1.6939039087447026e-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": 0.006295015336945653, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0028345256694592537, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.015424556471407414, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.01645718878135085, + "weighted_rkd_loss": 0.007712278235703707 + }, + { + "epoch": 4.093959731543624, + "flow_matching_loss": 0.019854934979230167, + "grad_norm": 0.21733368933200836, + "learning_rate": 1.6840621609248397e-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": 0.0071131703443825245, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0028163781855255364, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.017042718548327684, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.019854934979230167, + "weighted_rkd_loss": 0.008521359274163842 + }, + { + "epoch": 4.110738255033557, + "flow_matching_loss": 0.015388017985969782, + "grad_norm": 0.20185643434524536, + "learning_rate": 1.6742123616297794e-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": 0.006749651581048966, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.002803330810274929, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.016302633937448262, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.015388017985969782, + "weighted_rkd_loss": 0.008151316968724131 + }, + { + "epoch": 4.12751677852349, + "flow_matching_loss": 0.016323147248476742, + "grad_norm": 0.1591213494539261, + "learning_rate": 1.664354941721651e-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": 0.006030449504032731, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0027405192493461074, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.014801418222486974, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.016323147248476742, + "weighted_rkd_loss": 0.007400709111243487 + }, + { + "epoch": 4.144295302013423, + "flow_matching_loss": 0.014793385285884142, + "grad_norm": 0.15707005560398102, + "learning_rate": 1.6544903323959355e-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": 0.005447660619392991, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.001981837511993945, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.012877158727496863, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.014793385285884142, + "weighted_rkd_loss": 0.006438579363748431 + }, + { + "epoch": 4.1610738255033555, + "flow_matching_loss": 0.018846805114299058, + "grad_norm": 0.24046272039413452, + "learning_rate": 1.6446189651626023e-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": 0.006962421163916588, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0027715746778994798, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.0166964172385633, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.018846805114299058, + "weighted_rkd_loss": 0.00834820861928165 + }, + { + "epoch": 4.177852348993288, + "flow_matching_loss": 0.01845680521801114, + "grad_norm": 0.20694294571876526, + "learning_rate": 1.6347412718272316e-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": 0.0066049370681867, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.002969217847567052, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.016179092042148114, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.01845680521801114, + "weighted_rkd_loss": 0.008089546021074057 + }, + { + "epoch": 4.194630872483222, + "flow_matching_loss": 0.01491537420079112, + "grad_norm": 0.20724180340766907, + "learning_rate": 1.624857684472131e-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": 0.006738457502797246, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.003025981585960835, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.016502896696329115, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.01491537420079112, + "weighted_rkd_loss": 0.008251448348164558 + }, + { + "epoch": 4.2114093959731544, + "flow_matching_loss": 0.016149316169321536, + "grad_norm": 0.1734648197889328, + "learning_rate": 1.6149686354374283e-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": 0.006363571970723569, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0025395003031007946, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.015266644209623337, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.016149316169321536, + "weighted_rkd_loss": 0.007633322104811669 + }, + { + "epoch": 4.228187919463087, + "flow_matching_loss": 0.014004068076610565, + "grad_norm": 0.23745885491371155, + "learning_rate": 1.605074557302165e-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": 0.006188544281758368, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0025183343095704913, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.014895422849804163, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.014004068076610565, + "weighted_rkd_loss": 0.0074477114249020815 + }, + { + "epoch": 4.24496644295302, + "flow_matching_loss": 0.019259376171976327, + "grad_norm": 0.21155360341072083, + "learning_rate": 1.59517588286537e-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": 0.006474639987573028, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.002942153089679778, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.015891432855278255, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.019259376171976327, + "weighted_rkd_loss": 0.007945716427639127 + }, + { + "epoch": 4.261744966442953, + "flow_matching_loss": 0.01801548944786191, + "grad_norm": 0.20064517855644226, + "learning_rate": 1.5852730451271324e-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": 0.005549517436884343, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.002522014582064003, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.013621049467474222, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.01801548944786191, + "weighted_rkd_loss": 0.006810524733737111 + }, + { + "epoch": 4.278523489932886, + "flow_matching_loss": 0.015501426067203283, + "grad_norm": 0.20166797935962677, + "learning_rate": 1.5753664772696546e-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": 0.0064768324140459296, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0024600263452157376, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.015413691382855177, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.015501426067203283, + "weighted_rkd_loss": 0.007706845691427588 + }, + { + "epoch": 4.295302013422819, + "flow_matching_loss": 0.015355982910841703, + "grad_norm": 0.17986415326595306, + "learning_rate": 1.565456612638307e-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": 0.005972770554944873, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0026128120254725216, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.014558353181928396, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.015355982910841703, + "weighted_rkd_loss": 0.007279176590964198 + }, + { + "epoch": 4.3120805369127515, + "flow_matching_loss": 0.017460724618285895, + "grad_norm": 0.1711965799331665, + "learning_rate": 1.5555438847226732e-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": 0.00691812839359045, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.002547812880948186, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.016384069714695214, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.017460724618285895, + "weighted_rkd_loss": 0.008192034857347607 + }, + { + "epoch": 4.328859060402684, + "flow_matching_loss": 0.01676307516172528, + "grad_norm": 0.20304787158966064, + "learning_rate": 1.545628727137584e-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": 0.008595090778544544, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0034613806754350664, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.020651562511920928, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.01676307516172528, + "weighted_rkd_loss": 0.010325781255960464 + }, + { + "epoch": 4.345637583892618, + "flow_matching_loss": 0.017132134642452, + "grad_norm": 0.2237437218427658, + "learning_rate": 1.535711573604153e-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": 0.007427459000609815, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.003046630509197712, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.0179015482775867, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.017132134642452, + "weighted_rkd_loss": 0.00895077413879335 + }, + { + "epoch": 4.3624161073825505, + "flow_matching_loss": 0.015688648354262112, + "grad_norm": 0.15388670563697815, + "learning_rate": 1.5257928579308032e-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": 0.0060683083022013305, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0025359545601531864, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.014672570955008268, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.015688648354262112, + "weighted_rkd_loss": 0.007336285477504134 + }, + { + "epoch": 4.379194630872483, + "flow_matching_loss": 0.016403859294950963, + "grad_norm": 0.2258034348487854, + "learning_rate": 1.5158730139942902e-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": 0.007638280279934406, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0033454284304752944, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.018621988873928786, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.016403859294950963, + "weighted_rkd_loss": 0.009310994436964393 + }, + { + "epoch": 4.395973154362416, + "flow_matching_loss": 0.018862981349229813, + "grad_norm": 0.1582634150981903, + "learning_rate": 1.5059524757207239e-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": 0.005366501468233764, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.002398842200636864, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.013131845183670521, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.018862981349229813, + "weighted_rkd_loss": 0.0065659225918352606 + }, + { + "epoch": 4.412751677852349, + "flow_matching_loss": 0.015344698913395405, + "grad_norm": 0.17611289024353027, + "learning_rate": 1.4960316770665866e-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": 0.006546007725410164, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0035214671748690306, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.016613482777029277, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.015344698913395405, + "weighted_rkd_loss": 0.008306741388514638 + }, + { + "epoch": 4.429530201342282, + "flow_matching_loss": 0.015177258756011724, + "grad_norm": 0.1834149956703186, + "learning_rate": 1.4861110519997495e-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": 0.0069353858474642035, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0028723087976686656, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01674308059737086, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.015177258756011724, + "weighted_rkd_loss": 0.00837154029868543 + }, + { + "epoch": 4.446308724832215, + "flow_matching_loss": 0.015033644996583462, + "grad_norm": 0.19390572607517242, + "learning_rate": 1.4761910344804913e-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": 0.005982156842947006, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0027972670504823327, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.014761580899357796, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.015033644996583462, + "weighted_rkd_loss": 0.007380790449678898 + }, + { + "epoch": 4.4630872483221475, + "flow_matching_loss": 0.017534072417765854, + "grad_norm": 0.15612713992595673, + "learning_rate": 1.466272058442516e-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": 0.006262251944281161, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0026205297675915063, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01514503387734294, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.017534072417765854, + "weighted_rkd_loss": 0.00757251693867147 + }, + { + "epoch": 4.47986577181208, + "flow_matching_loss": 0.01529231872409582, + "grad_norm": 0.17678481340408325, + "learning_rate": 1.4563545577739676e-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": 0.0058165386784821745, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0026954136905260385, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.014328491035848856, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.01529231872409582, + "weighted_rkd_loss": 0.007164245517924428 + }, + { + "epoch": 4.496644295302014, + "flow_matching_loss": 0.013704319205135106, + "grad_norm": 0.1768968105316162, + "learning_rate": 1.4464389662984543e-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": 0.007249598205089569, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.003044455614872277, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01754365172237158, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.013704319205135106, + "weighted_rkd_loss": 0.00877182586118579 + }, + { + "epoch": 4.5134228187919465, + "flow_matching_loss": 0.014010507240891456, + "grad_norm": 0.2133304327726364, + "learning_rate": 1.4365257177560684e-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": 0.006245337147265673, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0024208846618421374, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.014911558851599694, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.014010507240891456, + "weighted_rkd_loss": 0.007455779425799847 + }, + { + "epoch": 4.530201342281879, + "flow_matching_loss": 0.01742522157728672, + "grad_norm": 0.1453578770160675, + "learning_rate": 1.426615245784416e-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": 0.006315420614555478, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.002484260895289481, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.015115102287381887, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.01742522157728672, + "weighted_rkd_loss": 0.007557551143690944 + }, + { + "epoch": 4.546979865771812, + "flow_matching_loss": 0.014088719990104437, + "grad_norm": 0.1421394795179367, + "learning_rate": 1.4167079838996468e-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": 0.0060806600376963615, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0026823743479326367, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.014843694493174553, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.014088719990104437, + "weighted_rkd_loss": 0.007421847246587276 + }, + { + "epoch": 4.563758389261745, + "flow_matching_loss": 0.014640630967915059, + "grad_norm": 0.18813489377498627, + "learning_rate": 1.4068043654774897e-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": 0.006225122767500579, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0027075539575889705, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01515779928304255, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.014640630967915059, + "weighted_rkd_loss": 0.007578899641521275 + }, + { + "epoch": 4.580536912751678, + "flow_matching_loss": 0.015772221237421037, + "grad_norm": 0.19396649301052094, + "learning_rate": 1.3969048237342992e-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": 0.006218162784352899, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.002695493830833584, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.015131819341331721, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.015772221237421037, + "weighted_rkd_loss": 0.0075659096706658605 + }, + { + "epoch": 4.597315436241611, + "flow_matching_loss": 0.017749189864844084, + "grad_norm": 0.3539927005767822, + "learning_rate": 1.3870097917081011e-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": 0.006628911755979061, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0026855499716475608, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.015943373460322617, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.017749189864844084, + "weighted_rkd_loss": 0.007971686730161309 + }, + { + "epoch": 4.614093959731544, + "flow_matching_loss": 0.016945687308907508, + "grad_norm": 0.19812190532684326, + "learning_rate": 1.3771197022396508e-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": 0.0064422482624650005, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.00251772936899215, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.015402225963771343, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.016945687308907508, + "weighted_rkd_loss": 0.007701112981885671 + }, + { + "epoch": 4.630872483221476, + "flow_matching_loss": 0.015047456696629524, + "grad_norm": 0.22193585336208344, + "learning_rate": 1.3672349879535015e-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": 0.006571926502510905, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.002829372195992619, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.015973225142806767, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.015047456696629524, + "weighted_rkd_loss": 0.007986612571403384 + }, + { + "epoch": 4.64765100671141, + "flow_matching_loss": 0.01549183800816536, + "grad_norm": 0.21233487129211426, + "learning_rate": 1.3573560812390781e-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": 0.006229676026850939, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.002528882946353406, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.014988235104829073, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.01549183800816536, + "weighted_rkd_loss": 0.007494117552414537 + }, + { + "epoch": 4.6644295302013425, + "flow_matching_loss": 0.01588299232535064, + "grad_norm": 0.15140114724636078, + "learning_rate": 1.3474834142317635e-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": 0.006161411851644516, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.002644290472380817, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.014967114012688399, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.01588299232535064, + "weighted_rkd_loss": 0.0074835570063441995 + }, + { + "epoch": 4.681208053691275, + "flow_matching_loss": 0.018512141052633525, + "grad_norm": 0.19018064439296722, + "learning_rate": 1.3376174187939955e-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": 0.0057833738625049595, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.002715141198132187, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.014281888864934445, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.018512141052633525, + "weighted_rkd_loss": 0.007140944432467222 + }, + { + "epoch": 4.697986577181208, + "flow_matching_loss": 0.01736147440969944, + "grad_norm": 0.1672631949186325, + "learning_rate": 1.3277585264963761e-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": 0.006036809016950429, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0023804905940778553, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.014454108569771052, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.01736147440969944, + "weighted_rkd_loss": 0.007227054284885526 + }, + { + "epoch": 4.714765100671141, + "flow_matching_loss": 0.014889136515557766, + "grad_norm": 0.2063864916563034, + "learning_rate": 1.3179071685987928e-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": 0.00579841728322208, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0028218236402608454, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.014418658148497344, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.014889136515557766, + "weighted_rkd_loss": 0.007209329074248672 + }, + { + "epoch": 4.731543624161074, + "flow_matching_loss": 0.014857401140034199, + "grad_norm": 0.19917383790016174, + "learning_rate": 1.308063776031554e-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": 0.00553174801170826, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0023466175771318376, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.013410113565623761, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.014857401140034199, + "weighted_rkd_loss": 0.0067050567828118805 + }, + { + "epoch": 4.748322147651007, + "flow_matching_loss": 0.015048641338944435, + "grad_norm": 0.17118246853351593, + "learning_rate": 1.2982287793765389e-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": 0.005542109534144401, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0022684723138809205, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.013352691382169723, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.015048641338944435, + "weighted_rkd_loss": 0.006676345691084861 + }, + { + "epoch": 4.76510067114094, + "flow_matching_loss": 0.01329810954630375, + "grad_norm": 0.17175836861133575, + "learning_rate": 1.2884026088483616e-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": 0.005096558784134686, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0020838214200921356, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.012276938930153847, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.01329810954630375, + "weighted_rkd_loss": 0.006138469465076924 + }, + { + "epoch": 4.781879194630872, + "flow_matching_loss": 0.01690172664821148, + "grad_norm": 0.18863530457019806, + "learning_rate": 1.2785856942755543e-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": 0.0061559075955301525, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.002671601914335042, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.014983417373150587, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.01690172664821148, + "weighted_rkd_loss": 0.007491708686575294 + }, + { + "epoch": 4.798657718120805, + "flow_matching_loss": 0.017869122605770827, + "grad_norm": 0.1789858490228653, + "learning_rate": 1.2687784650817616e-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": 0.005891825188882649, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0027419522521086037, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.014525602664798498, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.017869122605770827, + "weighted_rkd_loss": 0.007262801332399249 + }, + { + "epoch": 4.815436241610739, + "flow_matching_loss": 0.017372965905815364, + "grad_norm": 0.26185187697410583, + "learning_rate": 1.2589813502669603e-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": 0.006301722163334489, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0025509233120828865, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.015154367499053478, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.017372965905815364, + "weighted_rkd_loss": 0.007577183749526739 + }, + { + "epoch": 4.832214765100671, + "flow_matching_loss": 0.015076496172696352, + "grad_norm": 0.17579181492328644, + "learning_rate": 1.2491947783886907e-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": 0.004967443551868201, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0022041501477360727, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01213903734460473, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.015076496172696352, + "weighted_rkd_loss": 0.006069518672302365 + }, + { + "epoch": 4.848993288590604, + "flow_matching_loss": 0.015099618025124073, + "grad_norm": 0.16331081092357635, + "learning_rate": 1.2394191775433106e-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": 0.005304206721484661, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0022994426894001664, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.012907856423407794, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.015099618025124073, + "weighted_rkd_loss": 0.006453928211703897 + }, + { + "epoch": 4.865771812080537, + "flow_matching_loss": 0.014436965901404619, + "grad_norm": 0.2094336897134781, + "learning_rate": 1.2296549753472696e-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": 0.005188536667264998, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.002317892329301685, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.012694965675473213, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.014436965901404619, + "weighted_rkd_loss": 0.006347482837736606 + }, + { + "epoch": 4.882550335570469, + "flow_matching_loss": 0.01326000913977623, + "grad_norm": 0.21835879981517792, + "learning_rate": 1.2199025989184032e-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": 0.0054028286598622795, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0023249278543516993, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01313058529049158, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.01326000913977623, + "weighted_rkd_loss": 0.00656529264524579 + }, + { + "epoch": 4.899328859060403, + "flow_matching_loss": 0.013344433438032866, + "grad_norm": 0.19947674870491028, + "learning_rate": 1.21016247485725e-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": 0.006551743950694799, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.002417468442581594, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.015520956460386514, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.013344433438032866, + "weighted_rkd_loss": 0.007760478230193257 + }, + { + "epoch": 4.916107382550336, + "flow_matching_loss": 0.015554362256079913, + "grad_norm": 0.20881305634975433, + "learning_rate": 1.2004350292283897e-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": 0.006303047342225909, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.002661587006878108, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.015267681423574686, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.015554362256079913, + "weighted_rkd_loss": 0.007633840711787343 + }, + { + "epoch": 4.932885906040268, + "flow_matching_loss": 0.01382281593978405, + "grad_norm": 0.15891192853450775, + "learning_rate": 1.1907206875418068e-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": 0.005367122218012809, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0020784522290341555, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.012812696723267436, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.01382281593978405, + "weighted_rkd_loss": 0.006406348361633718 + }, + { + "epoch": 4.949664429530201, + "flow_matching_loss": 0.0140784309245646, + "grad_norm": 0.15928204357624054, + "learning_rate": 1.1810198747342769e-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": 0.004789382079616189, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.001973238063510507, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.011552002094686032, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.0140784309245646, + "weighted_rkd_loss": 0.005776001047343016 + }, + { + "epoch": 4.966442953020135, + "flow_matching_loss": 0.013056613504886627, + "grad_norm": 0.17947493493556976, + "learning_rate": 1.1713330151507786e-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": 0.005986737413331866, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.002562668244354427, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.014536143001168966, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.013056613504886627, + "weighted_rkd_loss": 0.007268071500584483 + }, + { + "epoch": 4.983221476510067, + "flow_matching_loss": 0.01408508289605379, + "grad_norm": 0.15849687159061432, + "learning_rate": 1.161660532525931e-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": 0.005506622069515288, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0025691559188999236, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01358239995315671, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.01408508289605379, + "weighted_rkd_loss": 0.006791199976578355 + }, + { + "epoch": 5.0, + "flow_matching_loss": 0.01281053079292178, + "grad_norm": 0.17440581321716309, + "learning_rate": 1.1520028499654584e-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": 0.005511744110845029, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0020654156105592847, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.013088903855532408, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.01281053079292178, + "weighted_rkd_loss": 0.006544451927766204 + }, + { + "epoch": 5.016778523489933, + "flow_matching_loss": 0.01506503028795123, + "grad_norm": 0.16012804210186005, + "learning_rate": 1.1423603899276833e-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": 0.005214577075093985, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.002262819651514292, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.012691974081099034, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.01506503028795123, + "weighted_rkd_loss": 0.006345987040549517 + }, + { + "epoch": 5.033557046979865, + "flow_matching_loss": 0.014499222207814456, + "grad_norm": 0.1774403154850006, + "learning_rate": 1.1327335742050442e-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": 0.005512195383198559, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0024933496373705567, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.013517740368843078, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.014499222207814456, + "weighted_rkd_loss": 0.006758870184421539 + } + ], + "logging_steps": 10, + "max_steps": 5000, + "num_input_tokens_seen": 0, + "num_train_epochs": 9, + "save_steps": 5001, + "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": 64, + "trial_name": null, + "trial_params": null +} diff --git a/FewShot/39tasks/5shot/FromRKD-Timewarp_FullactionHead/seed0/default/train_5shot_seed0_fullactionhead_5k/experiment_cfg/metadata.json b/FewShot/39tasks/5shot/FromRKD-Timewarp_FullactionHead/seed0/default/train_5shot_seed0_fullactionhead_5k/experiment_cfg/metadata.json new file mode 100644 index 0000000000000000000000000000000000000000..6f73f8489294c58acc82d1351a17e5e3ff316e18 --- /dev/null +++ b/FewShot/39tasks/5shot/FromRKD-Timewarp_FullactionHead/seed0/default/train_5shot_seed0_fullactionhead_5k/experiment_cfg/metadata.json @@ -0,0 +1,431 @@ +{ + "new_embodiment": { + "statistics": { + "state": { + "base_position": { + "max": [ + 7.778976917266846, + 0.33403685688972473, + 0.7223060727119446 + ], + "min": [ + -4.821934223175049, + -6.785110950469971, + 0.6994727253913879 + ], + "mean": [ + 2.5886094736594543, + -2.1559664577316586, + 0.7003580854548973 + ], + "std": [ + 1.6717149343420803, + 1.3743695399555862, + 0.000940069542359506 + ], + "q01": [ + -0.7517815060249613, + -5.497849509089705, + 0.6999776914424374 + ], + "q99": [ + 6.060945690807028, + -0.5331499932493624, + 0.7028864260472257 + ] + }, + "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, + 0.24488939171811902, + 0.5968549204077488 + ], + "std": [ + 0.0, + 0.0, + 0.6659970734507764, + 0.37443298392262264 + ], + "q01": [ + 0.0, + 0.0, + -0.9999999999999999, + 4.149924358336069e-06 + ], + "q99": [ + 0.0, + 0.0, + 0.9999999999999999, + 0.9999999999999999 + ] + }, + "end_effector_position_relative": { + "max": [ + 0.8848381638526917, + 0.8569459915161133, + 1.053884744644165 + ], + "min": [ + -0.5828549861907959, + -0.8486236333847046, + -0.3446122705936432 + ], + "mean": [ + 0.25087486209693133, + -0.005738920462036393, + 0.43643948866361787 + ], + "std": [ + 0.15741544532518012, + 0.2547603106793105, + 0.2050434492495558 + ], + "q01": [ + -0.013990057288082561, + -0.4824599092623036, + 0.18825220353500022 + ], + "q99": [ + 0.5411989642841469, + 0.4754923657245658, + 0.6912323747882034 + ] + }, + "end_effector_rotation_relative": { + "max": [ + 0.9999949932098389, + 0.9997354745864868, + 0.9900747537612915, + 0.9735639095306396 + ], + "min": [ + -0.9999914169311523, + -0.9999884366989136, + -0.9975369572639465, + 6.754255821306288e-08 + ], + "mean": [ + -0.22290331318740714, + 0.04579037037503981, + -0.05804094906527823, + 0.12347219024917105 + ], + "std": [ + 0.8519267931761344, + 0.2949394732270141, + 0.30026078832556913, + 0.16333086331829588 + ], + "q01": [ + -0.9957959450418518, + -0.5678869493126506, + -0.5318135475761875, + 0.0014160465686576505 + ], + "q99": [ + 0.9900692296898317, + 0.6357358195673684, + 0.5009847868015971, + 0.4287044482210845 + ] + }, + "gripper_qpos": { + "max": [ + 0.05289580300450325, + 0.00937709677964449 + ], + "min": [ + -0.011669117026031017, + -0.06307756155729294 + ], + "mean": [ + 0.03203268930378034, + -0.032331639271968075 + ], + "std": [ + 0.011607049173153697, + 0.011306286745872529 + ], + "q01": [ + 0.006429109393656672, + -0.04056586943958638 + ], + "q99": [ + 0.040377860142451545, + -0.006593889698673258 + ] + } + }, + "action": { + "base_motion": { + "max": [ + 1.0, + 1.0, + 1.0, + 0.0 + ], + "min": [ + -1.0, + -1.0, + -1.0, + 0.0 + ], + "mean": [ + 0.006050336852760097, + 0.002728916466628117, + -0.0017236811393598838, + 0.0 + ], + "std": [ + 0.11032167711449951, + 0.09824140301615687, + 0.07825208047515658, + 0.0 + ], + "q01": [ + -0.154721262094005, + -0.18955302988697728, + -0.14454065867754942, + 0.0 + ], + "q99": [ + 0.24492955247744197, + 0.2496797822190429, + 0.11966742325336034, + 0.0 + ] + }, + "control_mode": { + "max": [ + 1.0 + ], + "min": [ + -1.0 + ], + "mean": [ + -0.9294448715752568 + ], + "std": [ + 0.36897903503358326 + ], + "q01": [ + -1.0 + ], + "q99": [ + -0.21627562513311385 + ] + }, + "end_effector_position": { + "max": [ + 1.0, + 1.0, + 1.0 + ], + "min": [ + -1.0, + -1.0, + -1.0 + ], + "mean": [ + -0.00826780165926883, + 0.003161997359379232, + -0.08157453651841007 + ], + "std": [ + 0.4545430944464769, + 0.41040151434986594, + 0.4102644251536544 + ], + "q01": [ + -0.9653185255999107, + -0.9255635199318754, + -0.8208437313341718 + ], + "q99": [ + 0.7987781084592441, + 0.9260844232330088, + 0.8357421812622201 + ] + }, + "end_effector_rotation": { + "max": [ + 1.0, + 1.0, + 1.0 + ], + "min": [ + -1.0, + -1.0, + -1.0 + ], + "mean": [ + 0.007367024641295137, + -0.02060756343728361, + -0.001884833075998489 + ], + "std": [ + 0.1116768956941076, + 0.13249042272165212, + 0.12338908364725322 + ], + "q01": [ + -0.26894658663246845, + -0.39865607740339326, + -0.34393735191302655 + ], + "q99": [ + 0.3191674780171587, + 0.31133630397431267, + 0.33968476423051097 + ] + }, + "gripper_close": { + "max": [ + 1.0 + ], + "min": [ + -1.0 + ], + "mean": [ + -0.3098994894722959 + ], + "std": [ + 0.9507718287498876 + ], + "q01": [ + -1.0 + ], + "q99": [ + 0.7545620899936146 + ] + } + } + }, + "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/FewShot/39tasks/5shot/FromRKD-Timewarp_FullactionHead/seed1/default/resolved_config.yaml b/FewShot/39tasks/5shot/FromRKD-Timewarp_FullactionHead/seed1/default/resolved_config.yaml new file mode 100644 index 0000000000000000000000000000000000000000..05583ac4a8d843507d29ec548a5a6d86dfcc8386 --- /dev/null +++ b/FewShot/39tasks/5shot/FromRKD-Timewarp_FullactionHead/seed1/default/resolved_config.yaml @@ -0,0 +1,38 @@ +experiment_name: fewshot_39tasks_5shot_from_rkd_timewarp_fullactionhead_seed1 +policy_type: groot_RKDv3 +base_model_path: ${HOME}/groot_robocasa/robocasa_v2/TimewarpVAE/RKDv3_DA_APHead_from_Atomic26Baseline60k/default/phase3 +output_root: ${HOME}/groot_robocasa/robocasa_v2/FewShot/39tasks/5shot/FromRKD-Timewarp_FullactionHead/seed1 +dataset: + dataset_soup: fewshot_heldout_atomic39_human_5shot +training: + num_gpus: 1 + batch_size: 64 + seed: 42 + filter_key_seed: 1 + manifest_path: ${HOME}/clvla/benchmarks/robocasa_v2/my_scripts/Fewshot/manifests/fewshot_heldout_atomic39/shot5_seed1.json + save_total_limit: 4 +model: + rkd_enabled: true + rkd_fm_loss_weight: 1.0 + rkd_loss_weight: 0.5 + rkd_loss_weight_schedule: null +phases: +- name: train_5shot_seed1_fullactionhead_5k + max_steps: 5000 + save_steps: 0 + save_step_list: + - 1000 + - 3000 + - 5000 + save_final_model: false + 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: 0.5 + rkd_loss_weight_schedule: null +resolved_sweep: {} diff --git a/FewShot/39tasks/5shot/FromRKD-Timewarp_FullactionHead/seed1/default/seed1.yaml b/FewShot/39tasks/5shot/FromRKD-Timewarp_FullactionHead/seed1/default/seed1.yaml new file mode 100644 index 0000000000000000000000000000000000000000..fcaca05a1ddc4536d7ab79b3b76a28854f854627 --- /dev/null +++ b/FewShot/39tasks/5shot/FromRKD-Timewarp_FullactionHead/seed1/default/seed1.yaml @@ -0,0 +1,37 @@ +experiment_name: fewshot_39tasks_5shot_from_rkd_timewarp_fullactionhead_seed1 +policy_type: groot_RKDv3 +base_model_path: ${HOME}/groot_robocasa/robocasa_v2/TimewarpVAE/RKDv3_DA_APHead_from_Atomic26Baseline60k/default/phase3 +output_root: ${HOME}/groot_robocasa/robocasa_v2/FewShot/39tasks/5shot/FromRKD-Timewarp_FullactionHead/seed1 +dataset: + dataset_soup: fewshot_heldout_atomic39_human_5shot +training: + num_gpus: 1 + batch_size: 64 + seed: 42 + filter_key_seed: 1 + manifest_path: ${HOME}/clvla/benchmarks/robocasa_v2/my_scripts/Fewshot/manifests/fewshot_heldout_atomic39/shot5_seed1.json + save_total_limit: 4 +model: + rkd_enabled: true + rkd_fm_loss_weight: 1.0 + rkd_loss_weight: 0.5 + rkd_loss_weight_schedule: null +phases: +- name: train_5shot_seed1_fullactionhead_5k + max_steps: 5000 + save_steps: 0 + save_step_list: + - 1000 + - 3000 + - 5000 + save_final_model: false + 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: 0.5 + rkd_loss_weight_schedule: null diff --git a/FewShot/39tasks/5shot/FromRKD-Timewarp_FullactionHead/seed1/default/train_5shot_seed1_fullactionhead_5k/checkpoint-1000/config.json b/FewShot/39tasks/5shot/FromRKD-Timewarp_FullactionHead/seed1/default/train_5shot_seed1_fullactionhead_5k/checkpoint-1000/config.json new file mode 100644 index 0000000000000000000000000000000000000000..4740d0455bef31c469ee5dab6d4a47d64df61665 --- /dev/null +++ b/FewShot/39tasks/5shot/FromRKD-Timewarp_FullactionHead/seed1/default/train_5shot_seed1_fullactionhead_5k/checkpoint-1000/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": 0.5, + "rkd_loss_weight_end": 0.0, + "rkd_loss_weight_schedule": null, + "rkd_loss_weight_start": 0.1, + "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/FewShot/39tasks/5shot/FromRKD-Timewarp_FullactionHead/seed1/default/train_5shot_seed1_fullactionhead_5k/checkpoint-1000/experiment_cfg/metadata.json b/FewShot/39tasks/5shot/FromRKD-Timewarp_FullactionHead/seed1/default/train_5shot_seed1_fullactionhead_5k/checkpoint-1000/experiment_cfg/metadata.json new file mode 100644 index 0000000000000000000000000000000000000000..27700cac5d292d4d42897b3c2e80bdc58e55c60e --- /dev/null +++ b/FewShot/39tasks/5shot/FromRKD-Timewarp_FullactionHead/seed1/default/train_5shot_seed1_fullactionhead_5k/checkpoint-1000/experiment_cfg/metadata.json @@ -0,0 +1,431 @@ +{ + "new_embodiment": { + "statistics": { + "state": { + "base_position": { + "max": [ + 7.778976917266846, + 0.33403685688972473, + 0.7223060727119446 + ], + "min": [ + -4.821934223175049, + -6.785110950469971, + 0.6994727253913879 + ], + "mean": [ + 2.5719731235192254, + -2.1489371002267235, + 0.7003612741268546 + ], + "std": [ + 1.669945884010784, + 1.3739411381361726, + 0.0009413282367754987 + ], + "q01": [ + -0.7944286339399775, + -5.486611518471242, + 0.699977555971739 + ], + "q99": [ + 6.019886905263648, + -0.5229184210026855, + 0.702930326282967 + ] + }, + "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, + 0.248543484335273, + 0.5959317405459531 + ], + "std": [ + 0.0, + 0.0, + 0.6655550240074092, + 0.374283995143559 + ], + "q01": [ + 0.0, + 0.0, + -1.0000000000000002, + 4.503254968910557e-06 + ], + "q99": [ + 0.0, + 0.0, + 1.0000000000000002, + 1.0000000000000002 + ] + }, + "end_effector_position_relative": { + "max": [ + 0.8848381638526917, + 0.8569459915161133, + 1.053884744644165 + ], + "min": [ + -0.5828549861907959, + -0.8486236333847046, + -0.3446122705936432 + ], + "mean": [ + 0.2465490874130862, + -0.004711658908481467, + 0.43463835464312073 + ], + "std": [ + 0.15798784670671595, + 0.25624102400497073, + 0.20836028727974915 + ], + "q01": [ + -0.020752051134640608, + -0.4837933887347875, + 0.1802942952289915 + ], + "q99": [ + 0.5386784714844007, + 0.4801424212024721, + 0.695095732975766 + ] + }, + "end_effector_rotation_relative": { + "max": [ + 0.9999949932098389, + 0.9997354745864868, + 0.9900747537612915, + 0.9735639095306396 + ], + "min": [ + -0.9999914169311523, + -0.9999884366989136, + -0.9975369572639465, + 6.754255821306288e-08 + ], + "mean": [ + -0.22531859714056926, + 0.046973702821026614, + -0.0568176967878771, + 0.12364307042955697 + ], + "std": [ + 0.8502144333612052, + 0.29591250318147405, + 0.30218053156837016, + 0.1636031353412931 + ], + "q01": [ + -0.9959057387172566, + -0.5665384745573048, + -0.5329334620834977, + 0.0014372279999397553 + ], + "q99": [ + 0.9904107951565839, + 0.6359623612021396, + 0.5026076022439377, + 0.42955559234879914 + ] + }, + "gripper_qpos": { + "max": [ + 0.05289580300450325, + 0.00937709677964449 + ], + "min": [ + -0.011669117026031017, + -0.06307756155729294 + ], + "mean": [ + 0.03188757060429838, + -0.0321793333086053 + ], + "std": [ + 0.01171052313737148, + 0.011427336801203189 + ], + "q01": [ + 0.006462901319459925, + -0.040557140599419166 + ], + "q99": [ + 0.04037619335388386, + -0.006523947835993347 + ] + } + }, + "action": { + "base_motion": { + "max": [ + 1.0, + 1.0, + 1.0, + 0.0 + ], + "min": [ + -1.0, + -1.0, + -1.0, + 0.0 + ], + "mean": [ + 0.006639007355857881, + 0.0027330421795561874, + -0.0017774441970214892, + 0.0 + ], + "std": [ + 0.11841528114040736, + 0.1041359083946544, + 0.08346265521208351, + 0.0 + ], + "q01": [ + -0.17562359675270237, + -0.21706402321529558, + -0.16273392585723104, + 0.0 + ], + "q99": [ + 0.2572918701644887, + 0.27022325575787115, + 0.1369626751049278, + 0.0 + ] + }, + "control_mode": { + "max": [ + 1.0 + ], + "min": [ + -1.0 + ], + "mean": [ + -0.9205539608927732 + ], + "std": [ + 0.39063657242534666 + ], + "q01": [ + -1.0 + ], + "q99": [ + -0.17711863866562935 + ] + }, + "end_effector_position": { + "max": [ + 1.0, + 1.0, + 1.0 + ], + "min": [ + -1.0, + -1.0, + -1.0 + ], + "mean": [ + -0.00924803964621179, + 0.004053841761390575, + -0.08385115578460103 + ], + "std": [ + 0.45413072518223824, + 0.40158612714634856, + 0.41068634292849304 + ], + "q01": [ + -0.9669817145875343, + -0.9241307799783718, + -0.8264079503068825 + ], + "q99": [ + 0.8021511975722166, + 0.9241346582748478, + 0.8348116781502298 + ] + }, + "end_effector_rotation": { + "max": [ + 1.0, + 1.0, + 1.0 + ], + "min": [ + -1.0, + -1.0, + -1.0 + ], + "mean": [ + 0.007644157504353809, + -0.019962543093886335, + -0.002047110452602617 + ], + "std": [ + 0.11076012154791087, + 0.13176735140285795, + 0.12286528665895614 + ], + "q01": [ + -0.2659105439401192, + -0.3961694033813533, + -0.3434870832468936 + ], + "q99": [ + 0.3145786959266369, + 0.31309430361227547, + 0.34107754401942986 + ] + }, + "gripper_close": { + "max": [ + 1.0 + ], + "min": [ + -1.0 + ], + "mean": [ + -0.29870612267533575 + ], + "std": [ + 0.9543487555375691 + ], + "q01": [ + -1.0 + ], + "q99": [ + 0.7610674802484976 + ] + } + } + }, + "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/FewShot/39tasks/5shot/FromRKD-Timewarp_FullactionHead/seed1/default/train_5shot_seed1_fullactionhead_5k/checkpoint-1000/model.safetensors.index.json b/FewShot/39tasks/5shot/FromRKD-Timewarp_FullactionHead/seed1/default/train_5shot_seed1_fullactionhead_5k/checkpoint-1000/model.safetensors.index.json new file mode 100644 index 0000000000000000000000000000000000000000..b20baed71f4d16a411223b70d45d934c6cbbfe97 --- /dev/null +++ b/FewShot/39tasks/5shot/FromRKD-Timewarp_FullactionHead/seed1/default/train_5shot_seed1_fullactionhead_5k/checkpoint-1000/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/FewShot/39tasks/5shot/FromRKD-Timewarp_FullactionHead/seed1/default/train_5shot_seed1_fullactionhead_5k/checkpoint-1000/trainer_state.json b/FewShot/39tasks/5shot/FromRKD-Timewarp_FullactionHead/seed1/default/train_5shot_seed1_fullactionhead_5k/checkpoint-1000/trainer_state.json new file mode 100644 index 0000000000000000000000000000000000000000..940bccf93363d2ae8b86917e6a3063787a55d2af --- /dev/null +++ b/FewShot/39tasks/5shot/FromRKD-Timewarp_FullactionHead/seed1/default/train_5shot_seed1_fullactionhead_5k/checkpoint-1000/trainer_state.json @@ -0,0 +1,4634 @@ +{ + "best_global_step": null, + "best_metric": null, + "best_model_checkpoint": null, + "epoch": 1.675041876046901, + "eval_steps": 500, + "global_step": 1000, + "is_hyper_param_search": false, + "is_local_process_zero": true, + "is_world_process_zero": true, + "log_history": [ + { + "epoch": 0.01675041876046901, + "flow_matching_loss": 0.18734376579523088, + "grad_norm": 1.2580132484436035, + "learning_rate": 1.08e-06, + "loss": 0.384, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.13968379348516463, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.11389333754777908, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.3932609289884567, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.18734376579523088, + "weighted_rkd_loss": 0.19663046449422836 + }, + { + "epoch": 0.03350083752093802, + "flow_matching_loss": 0.19343584030866623, + "grad_norm": 0.9545837044715881, + "learning_rate": 2.28e-06, + "loss": 0.3738, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.12992414832115173, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.10088557079434395, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.3607338696718216, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.19343584030866623, + "weighted_rkd_loss": 0.1803669348359108 + }, + { + "epoch": 0.05025125628140704, + "flow_matching_loss": 0.16988817751407623, + "grad_norm": 0.9067039489746094, + "learning_rate": 3.48e-06, + "loss": 0.3561, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.13251580074429511, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.1073499672114849, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.37238157391548155, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.16988817751407623, + "weighted_rkd_loss": 0.18619078695774077 + }, + { + "epoch": 0.06700167504187604, + "flow_matching_loss": 0.1650318592786789, + "grad_norm": 1.1154892444610596, + "learning_rate": 4.68e-06, + "loss": 0.3387, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.12463015466928482, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.09805893376469613, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.34731924533843994, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.1650318592786789, + "weighted_rkd_loss": 0.17365962266921997 + }, + { + "epoch": 0.08375209380234507, + "flow_matching_loss": 0.13135033324360848, + "grad_norm": 0.7893933653831482, + "learning_rate": 5.8800000000000005e-06, + "loss": 0.2851, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.11209887266159058, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.08326855897903443, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.30746630430221555, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.13135033324360848, + "weighted_rkd_loss": 0.15373315215110778 + }, + { + "epoch": 0.10050251256281408, + "flow_matching_loss": 0.11830725595355034, + "grad_norm": 0.6967894434928894, + "learning_rate": 7.08e-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": 0.1122154638171196, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.08148682340979577, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.30591775476932526, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.11830725595355034, + "weighted_rkd_loss": 0.15295887738466263 + }, + { + "epoch": 0.11725293132328309, + "flow_matching_loss": 0.09991217777132988, + "grad_norm": 0.5718310475349426, + "learning_rate": 8.28e-06, + "loss": 0.2493, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.11076194941997528, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.07722590751945972, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.29874980747699736, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.09991217777132988, + "weighted_rkd_loss": 0.14937490373849868 + }, + { + "epoch": 0.13400335008375208, + "flow_matching_loss": 0.09349806271493435, + "grad_norm": 0.5672374367713928, + "learning_rate": 9.48e-06, + "loss": 0.232, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.10319361835718155, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.07060621418058873, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.2769934505224228, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.09349806271493435, + "weighted_rkd_loss": 0.1384967252612114 + }, + { + "epoch": 0.1507537688442211, + "flow_matching_loss": 0.08115713745355606, + "grad_norm": 0.5059905052185059, + "learning_rate": 1.068e-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": 0.09615755379199982, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0642927698791027, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.2566078782081604, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.08115713745355606, + "weighted_rkd_loss": 0.1283039391040802 + }, + { + "epoch": 0.16750418760469013, + "flow_matching_loss": 0.07341464199125766, + "grad_norm": 0.5954922437667847, + "learning_rate": 1.1880000000000001e-05, + "loss": 0.1941, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.09074362516403198, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.05983745716512203, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.24132470935583114, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.07341464199125766, + "weighted_rkd_loss": 0.12066235467791557 + }, + { + "epoch": 0.18425460636515914, + "flow_matching_loss": 0.06752067059278488, + "grad_norm": 0.53483647108078, + "learning_rate": 1.308e-05, + "loss": 0.1833, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.08770871832966805, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0561994768679142, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.23161691278219224, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.06752067059278488, + "weighted_rkd_loss": 0.11580845639109612 + }, + { + "epoch": 0.20100502512562815, + "flow_matching_loss": 0.0645824696868658, + "grad_norm": 0.4954916536808014, + "learning_rate": 1.428e-05, + "loss": 0.1694, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.08025030866265297, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.049181188270449636, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.20968180298805236, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.0645824696868658, + "weighted_rkd_loss": 0.10484090149402618 + }, + { + "epoch": 0.21775544388609716, + "flow_matching_loss": 0.06256739944219589, + "grad_norm": 0.6052863001823425, + "learning_rate": 1.548e-05, + "loss": 0.1603, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.07429649494588375, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.04693063050508499, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.19552361965179443, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.06256739944219589, + "weighted_rkd_loss": 0.09776180982589722 + }, + { + "epoch": 0.23450586264656617, + "flow_matching_loss": 0.0674729149788618, + "grad_norm": 0.6622745990753174, + "learning_rate": 1.6680000000000003e-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": 0.07380652725696564, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.04217908121645451, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.18979213535785674, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.0674729149788618, + "weighted_rkd_loss": 0.09489606767892837 + }, + { + "epoch": 0.25125628140703515, + "flow_matching_loss": 0.05902594663202763, + "grad_norm": 0.596757173538208, + "learning_rate": 1.7879999999999998e-05, + "loss": 0.1449, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.06658526808023453, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.03863737247884273, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.1718079075217247, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.05902594663202763, + "weighted_rkd_loss": 0.08590395376086235 + }, + { + "epoch": 0.26800670016750416, + "flow_matching_loss": 0.062078724056482314, + "grad_norm": 0.5115512609481812, + "learning_rate": 1.908e-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": 0.06346513852477073, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.03915274403989315, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.1660830244421959, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.062078724056482314, + "weighted_rkd_loss": 0.08304151222109794 + }, + { + "epoch": 0.2847571189279732, + "flow_matching_loss": 0.05774013251066208, + "grad_norm": 0.4801546335220337, + "learning_rate": 2.0280000000000002e-05, + "loss": 0.1378, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.061606602370738985, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.03699663132429123, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.1602098360657692, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.05774013251066208, + "weighted_rkd_loss": 0.0801049180328846 + }, + { + "epoch": 0.3015075376884422, + "flow_matching_loss": 0.05709497965872288, + "grad_norm": 0.41434067487716675, + "learning_rate": 2.148e-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": 0.0580575879663229, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0330570163205266, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.1491721898317337, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.05709497965872288, + "weighted_rkd_loss": 0.07458609491586685 + }, + { + "epoch": 0.31825795644891125, + "flow_matching_loss": 0.05621899552643299, + "grad_norm": 0.4601005017757416, + "learning_rate": 2.268e-05, + "loss": 0.1284, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.056104433909058574, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.03210239075124264, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.1443112574517727, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.05621899552643299, + "weighted_rkd_loss": 0.07215562872588635 + }, + { + "epoch": 0.33500837520938026, + "flow_matching_loss": 0.057572386041283605, + "grad_norm": 0.5096933245658875, + "learning_rate": 2.3880000000000002e-05, + "loss": 0.129, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.05570494718849659, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.03144418653100729, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.1428540788590908, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.057572386041283605, + "weighted_rkd_loss": 0.0714270394295454 + }, + { + "epoch": 0.35175879396984927, + "flow_matching_loss": 0.052765559032559396, + "grad_norm": 0.4708020091056824, + "learning_rate": 2.508e-05, + "loss": 0.1177, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.050074973702430726, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.029708896577358247, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.1298588439822197, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.052765559032559396, + "weighted_rkd_loss": 0.06492942199110985 + }, + { + "epoch": 0.3685092127303183, + "flow_matching_loss": 0.048277949541807176, + "grad_norm": 0.41520747542381287, + "learning_rate": 2.628e-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": 0.0443026889115572, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.024513646587729455, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.11311902478337288, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.048277949541807176, + "weighted_rkd_loss": 0.05655951239168644 + }, + { + "epoch": 0.3852596314907873, + "flow_matching_loss": 0.050647230073809626, + "grad_norm": 0.488421767950058, + "learning_rate": 2.748e-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": 0.044228341057896615, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.02473187427967787, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.11318855583667756, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.050647230073809626, + "weighted_rkd_loss": 0.05659427791833878 + }, + { + "epoch": 0.4020100502512563, + "flow_matching_loss": 0.052835336327552794, + "grad_norm": 0.4683481752872467, + "learning_rate": 2.868e-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": 0.043489795550704, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.025273922458291055, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.11225351244211197, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.052835336327552794, + "weighted_rkd_loss": 0.056126756221055986 + }, + { + "epoch": 0.4187604690117253, + "flow_matching_loss": 0.0462393905967474, + "grad_norm": 0.3933083415031433, + "learning_rate": 2.9880000000000002e-05, + "loss": 0.0989, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.04124091118574143, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.022835531365126373, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.10531735345721245, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.0462393905967474, + "weighted_rkd_loss": 0.052658676728606225 + }, + { + "epoch": 0.4355108877721943, + "flow_matching_loss": 0.04874549731612206, + "grad_norm": 0.40502575039863586, + "learning_rate": 2.9999734259635676e-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": 0.039895395189523696, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.021268937550485135, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.10105972811579704, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.04874549731612206, + "weighted_rkd_loss": 0.05052986405789852 + }, + { + "epoch": 0.45226130653266333, + "flow_matching_loss": 0.051396344974637034, + "grad_norm": 0.5639358162879944, + "learning_rate": 2.9998815663057245e-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": 0.03602162189781666, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.020294071175158024, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.09233731552958488, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.051396344974637034, + "weighted_rkd_loss": 0.04616865776479244 + }, + { + "epoch": 0.46901172529313234, + "flow_matching_loss": 0.04447287768125534, + "grad_norm": 0.40996530652046204, + "learning_rate": 2.999724096969228e-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": 0.0336500596255064, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.017429837398231028, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.08472995683550835, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.04447287768125534, + "weighted_rkd_loss": 0.042364978417754176 + }, + { + "epoch": 0.48576214405360135, + "flow_matching_loss": 0.04591554999351501, + "grad_norm": 0.4135409891605377, + "learning_rate": 2.9995010248422984e-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": 0.034760774858295915, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.018106953613460063, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.08762850314378738, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.04591554999351501, + "weighted_rkd_loss": 0.04381425157189369 + }, + { + "epoch": 0.5025125628140703, + "flow_matching_loss": 0.04205652140080929, + "grad_norm": 0.38845568895339966, + "learning_rate": 2.999212359682832e-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": 0.03268313799053431, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.017657354567199947, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.08302363082766533, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.04205652140080929, + "weighted_rkd_loss": 0.04151181541383266 + }, + { + "epoch": 0.5192629815745393, + "flow_matching_loss": 0.04226362742483616, + "grad_norm": 0.3450797200202942, + "learning_rate": 2.9988581141179787e-05, + "loss": 0.0805, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.030486975982785224, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.015418309066444636, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.07639226131141186, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.04226362742483616, + "weighted_rkd_loss": 0.03819613065570593 + }, + { + "epoch": 0.5360134003350083, + "flow_matching_loss": 0.0407985620200634, + "grad_norm": 0.3774641752243042, + "learning_rate": 2.998438303643587e-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": 0.03008116353303194, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.01613120548427105, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.07629353329539298, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.0407985620200634, + "weighted_rkd_loss": 0.03814676664769649 + }, + { + "epoch": 0.5527638190954773, + "flow_matching_loss": 0.041433697938919066, + "grad_norm": 0.37684324383735657, + "learning_rate": 2.997952946623527e-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": 0.03052981775254011, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.015478260163217784, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.07653789557516574, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.041433697938919066, + "weighted_rkd_loss": 0.03826894778758287 + }, + { + "epoch": 0.5695142378559463, + "flow_matching_loss": 0.04276835490018129, + "grad_norm": 0.41635242104530334, + "learning_rate": 2.9974020642888875e-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": 0.02872589062899351, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.014681840222328902, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.07213362231850624, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.04276835490018129, + "weighted_rkd_loss": 0.03606681115925312 + }, + { + "epoch": 0.5862646566164154, + "flow_matching_loss": 0.04388797134160995, + "grad_norm": 0.4484618604183197, + "learning_rate": 2.9967856807370458e-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": 0.03199950344860554, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.01809648834168911, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.08209549561142922, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.04388797134160995, + "weighted_rkd_loss": 0.04104774780571461 + }, + { + "epoch": 0.6030150753768844, + "flow_matching_loss": 0.04277344048023224, + "grad_norm": 0.4147380590438843, + "learning_rate": 2.996103822930615e-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": 0.029685858823359012, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.015408230014145374, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.07477994747459889, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.04277344048023224, + "weighted_rkd_loss": 0.037389973737299444 + }, + { + "epoch": 0.6197654941373534, + "flow_matching_loss": 0.04038205854594708, + "grad_norm": 0.371023952960968, + "learning_rate": 2.9953565206962653e-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": 0.028526164032518862, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.01493225758895278, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.07198458574712277, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.04038205854594708, + "weighted_rkd_loss": 0.035992292873561385 + }, + { + "epoch": 0.6365159128978225, + "flow_matching_loss": 0.04215538240969181, + "grad_norm": 0.3584204316139221, + "learning_rate": 2.9945438067234172e-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": 0.027064095251262188, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.014362859819084407, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.06849104985594749, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.04215538240969181, + "weighted_rkd_loss": 0.034245524927973746 + }, + { + "epoch": 0.6532663316582915, + "flow_matching_loss": 0.03905519917607307, + "grad_norm": 0.3805311620235443, + "learning_rate": 2.9936657165628124e-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": 0.026047029718756675, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.013608651608228684, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.06570271104574203, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.03905519917607307, + "weighted_rkd_loss": 0.032851355522871016 + }, + { + "epoch": 0.6700167504187605, + "flow_matching_loss": 0.03584225699305534, + "grad_norm": 0.35377806425094604, + "learning_rate": 2.992722288624959e-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": 0.02628317419439554, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.013779552932828665, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.06634590104222297, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.03584225699305534, + "weighted_rkd_loss": 0.03317295052111149 + }, + { + "epoch": 0.6867671691792295, + "flow_matching_loss": 0.03798154555261135, + "grad_norm": 0.33630454540252686, + "learning_rate": 2.9917135641784515e-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": 0.025117580965161323, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.013396832998842001, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.06363199502229691, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.03798154555261135, + "weighted_rkd_loss": 0.031815997511148456 + }, + { + "epoch": 0.7035175879396985, + "flow_matching_loss": 0.03501146622002125, + "grad_norm": 0.3093757629394531, + "learning_rate": 2.9906395873481655e-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": 0.024045810103416443, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.012114135175943374, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.06020575501024723, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.03501146622002125, + "weighted_rkd_loss": 0.030102877505123617 + }, + { + "epoch": 0.7202680067001676, + "flow_matching_loss": 0.038727116957306865, + "grad_norm": 0.2813211977481842, + "learning_rate": 2.989500405113326e-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": 0.025140311382710934, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.01257405923679471, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.0628546815365553, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.038727116957306865, + "weighted_rkd_loss": 0.03142734076827765 + }, + { + "epoch": 0.7370184254606366, + "flow_matching_loss": 0.035598259046673776, + "grad_norm": 0.3280915319919586, + "learning_rate": 2.9882960673054547e-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": 0.02384033761918545, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.012014199746772647, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.059694874286651614, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.035598259046673776, + "weighted_rkd_loss": 0.029847437143325807 + }, + { + "epoch": 0.7537688442211056, + "flow_matching_loss": 0.03603664711117745, + "grad_norm": 0.30604344606399536, + "learning_rate": 2.9870266266061886e-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": 0.022386168129742146, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.01189499134197831, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.05666732750833035, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.03603664711117745, + "weighted_rkd_loss": 0.028333663754165173 + }, + { + "epoch": 0.7705192629815746, + "flow_matching_loss": 0.036646503023803234, + "grad_norm": 0.3464244604110718, + "learning_rate": 2.9856921385449766e-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": 0.02474432121962309, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.012881245650351047, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.06236988790333271, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.036646503023803234, + "weighted_rkd_loss": 0.031184943951666357 + }, + { + "epoch": 0.7872696817420436, + "flow_matching_loss": 0.03805769421160221, + "grad_norm": 0.34830617904663086, + "learning_rate": 2.9842926614966502e-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": 0.02307860888540745, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.01193495150655508, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.05809217020869255, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.03805769421160221, + "weighted_rkd_loss": 0.029046085104346274 + }, + { + "epoch": 0.8040201005025126, + "flow_matching_loss": 0.03244946617633104, + "grad_norm": 0.36722618341445923, + "learning_rate": 2.982828256678868e-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": 0.02207295410335064, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.010983642470091582, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.05512955039739609, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.03244946617633104, + "weighted_rkd_loss": 0.027564775198698044 + }, + { + "epoch": 0.8207705192629816, + "flow_matching_loss": 0.035172288119792935, + "grad_norm": 0.2739450931549072, + "learning_rate": 2.9812989881494412e-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": 0.02087551765143871, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.010237869061529636, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.05198890343308449, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.035172288119792935, + "weighted_rkd_loss": 0.025994451716542244 + }, + { + "epoch": 0.8375209380234506, + "flow_matching_loss": 0.03737096916884184, + "grad_norm": 0.32612043619155884, + "learning_rate": 2.9797049228035296e-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": 0.02040377100929618, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.009668528474867344, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.05047607012093067, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.03737096916884184, + "weighted_rkd_loss": 0.025238035060465337 + }, + { + "epoch": 0.8542713567839196, + "flow_matching_loss": 0.03377289455384016, + "grad_norm": 0.29827985167503357, + "learning_rate": 2.978046130370715e-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": 0.019000360555946828, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.008991975570097566, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.04699269644916058, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.03377289455384016, + "weighted_rkd_loss": 0.02349634822458029 + }, + { + "epoch": 0.8710217755443886, + "flow_matching_loss": 0.034123093634843824, + "grad_norm": 0.2898612320423126, + "learning_rate": 2.9763226834119535e-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": 0.020308055356144904, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.010839158948510886, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.05145526938140392, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.034123093634843824, + "weighted_rkd_loss": 0.02572763469070196 + }, + { + "epoch": 0.8877721943048577, + "flow_matching_loss": 0.0334856515750289, + "grad_norm": 0.30531373620033264, + "learning_rate": 2.9745346573163963e-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": 0.022741097211837768, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.01176437777467072, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.05724657289683819, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.0334856515750289, + "weighted_rkd_loss": 0.028623286448419095 + }, + { + "epoch": 0.9045226130653267, + "flow_matching_loss": 0.03164167143404484, + "grad_norm": 0.38043874502182007, + "learning_rate": 2.9726821302980987e-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": 0.02146173259243369, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.010472412640228867, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.05339587815105915, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.03164167143404484, + "weighted_rkd_loss": 0.026697939075529576 + }, + { + "epoch": 0.9212730318257957, + "flow_matching_loss": 0.03139312732964754, + "grad_norm": 0.358161985874176, + "learning_rate": 2.9707651833925935e-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": 0.018665229622274637, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.00916362195275724, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.046494081243872644, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.03139312732964754, + "weighted_rkd_loss": 0.023247040621936322 + }, + { + "epoch": 0.9380234505862647, + "flow_matching_loss": 0.03471959326416254, + "grad_norm": 0.383441299200058, + "learning_rate": 2.968783900453348e-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": 0.02090346897020936, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.010127922613173723, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.05193486027419567, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.03471959326416254, + "weighted_rkd_loss": 0.025967430137097836 + }, + { + "epoch": 0.9547738693467337, + "flow_matching_loss": 0.03342412430793047, + "grad_norm": 0.3316456973552704, + "learning_rate": 2.9667383681480968e-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": 0.018697252310812474, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.00936074941419065, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.04675525352358818, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.03342412430793047, + "weighted_rkd_loss": 0.02337762676179409 + }, + { + "epoch": 0.9715242881072027, + "flow_matching_loss": 0.034992733597755434, + "grad_norm": 0.31459757685661316, + "learning_rate": 2.9646286759550493e-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": 0.019015263859182598, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.009189907228574157, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.04722043462097645, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.034992733597755434, + "weighted_rkd_loss": 0.023610217310488225 + }, + { + "epoch": 0.9882747068676717, + "flow_matching_loss": 0.03335383273661137, + "grad_norm": 0.39081528782844543, + "learning_rate": 2.9624549161589764e-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": 0.0166207866743207, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.007937656342983246, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.041179229691624644, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.03335383273661137, + "weighted_rkd_loss": 0.020589614845812322 + }, + { + "epoch": 1.0050251256281406, + "flow_matching_loss": 0.030267693661153318, + "grad_norm": 0.33321326971054077, + "learning_rate": 2.9602171838471745e-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": 0.02047300133854151, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.009730400750413537, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.05067640393972397, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.030267693661153318, + "weighted_rkd_loss": 0.025338201969861984 + }, + { + "epoch": 1.0217755443886096, + "flow_matching_loss": 0.028719464503228664, + "grad_norm": 0.33434903621673584, + "learning_rate": 2.957915576905303e-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": 0.01841264832764864, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.010385369556024671, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.047210666351020336, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.028719464503228664, + "weighted_rkd_loss": 0.023605333175510168 + }, + { + "epoch": 1.0385259631490786, + "flow_matching_loss": 0.029228216037154198, + "grad_norm": 0.31092560291290283, + "learning_rate": 2.9555501960131058e-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": 0.01791312750428915, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.008790831873193383, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.04461708702147007, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.029228216037154198, + "weighted_rkd_loss": 0.022308543510735034 + }, + { + "epoch": 1.0552763819095476, + "flow_matching_loss": 0.02980944626033306, + "grad_norm": 0.2826099097728729, + "learning_rate": 2.9531211446400068e-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": 0.018980990443378688, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.009706261800602078, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.04766824245452881, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.02980944626033306, + "weighted_rkd_loss": 0.023834121227264405 + }, + { + "epoch": 1.0720268006700167, + "flow_matching_loss": 0.030087723582983016, + "grad_norm": 0.3223755359649658, + "learning_rate": 2.9506285290405813e-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": 0.01794863734394312, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.00857453583739698, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.044471810758113864, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.030087723582983016, + "weighted_rkd_loss": 0.022235905379056932 + }, + { + "epoch": 1.0887772194304857, + "flow_matching_loss": 0.031088070012629032, + "grad_norm": 0.3262973725795746, + "learning_rate": 2.948072458249911e-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": 0.018507144786417483, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.008912034891545772, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.04592632427811623, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.031088070012629032, + "weighted_rkd_loss": 0.022963162139058114 + }, + { + "epoch": 1.1055276381909547, + "flow_matching_loss": 0.028495705127716063, + "grad_norm": 0.33219948410987854, + "learning_rate": 2.9454530440788125e-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": 0.02023590700700879, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.009690370690077544, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.0501621849834919, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.028495705127716063, + "weighted_rkd_loss": 0.02508109249174595 + }, + { + "epoch": 1.1222780569514237, + "flow_matching_loss": 0.03310795053839684, + "grad_norm": 0.3126451373100281, + "learning_rate": 2.9427704011089474e-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": 0.018809593096375466, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.008968799049034715, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.046587985008955, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.03310795053839684, + "weighted_rkd_loss": 0.0232939925044775 + }, + { + "epoch": 1.1390284757118927, + "flow_matching_loss": 0.03170393034815788, + "grad_norm": 0.275397390127182, + "learning_rate": 2.94002464668781e-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": 0.01641182703897357, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0077222629450261595, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.04054591674357653, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.03170393034815788, + "weighted_rkd_loss": 0.020272958371788264 + }, + { + "epoch": 1.1557788944723617, + "flow_matching_loss": 0.031644315272569654, + "grad_norm": 0.25714337825775146, + "learning_rate": 2.9372159009235936e-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": 0.017135247867554427, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.008725387183949352, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.04299588277935982, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.031644315272569654, + "weighted_rkd_loss": 0.02149794138967991 + }, + { + "epoch": 1.1725293132328307, + "flow_matching_loss": 0.03273461870849133, + "grad_norm": 0.22671352326869965, + "learning_rate": 2.9343442866799365e-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": 0.014938866905868053, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.006849234201945364, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.03672696817666292, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.03273461870849133, + "weighted_rkd_loss": 0.01836348408833146 + }, + { + "epoch": 1.1892797319932997, + "flow_matching_loss": 0.02922193594276905, + "grad_norm": 0.2870095670223236, + "learning_rate": 2.9314099295705495e-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": 0.016047394555062055, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.007527278270572424, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.039622067287564275, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.02922193594276905, + "weighted_rkd_loss": 0.019811033643782137 + }, + { + "epoch": 1.2060301507537687, + "flow_matching_loss": 0.026027432456612588, + "grad_norm": 0.22585110366344452, + "learning_rate": 2.928412957953718e-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": 0.014577559195458889, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.007318647485226393, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.03647376541048288, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.026027432456612588, + "weighted_rkd_loss": 0.01823688270524144 + }, + { + "epoch": 1.2227805695142377, + "flow_matching_loss": 0.029601436294615268, + "grad_norm": 0.337597131729126, + "learning_rate": 2.925353502926689e-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": 0.01586325457319617, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.007780421571806073, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.03950693104416132, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.029601436294615268, + "weighted_rkd_loss": 0.01975346552208066 + }, + { + "epoch": 1.2395309882747068, + "flow_matching_loss": 0.024688402004539967, + "grad_norm": 0.32554107904434204, + "learning_rate": 2.922231698319937e-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": 0.014900147635489702, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.006712513882666827, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.03651280924677849, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.024688402004539967, + "weighted_rkd_loss": 0.018256404623389243 + }, + { + "epoch": 1.2562814070351758, + "flow_matching_loss": 0.02895826380699873, + "grad_norm": 0.2865455746650696, + "learning_rate": 2.9190476806913088e-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": 0.014406697638332844, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.006247629038989544, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.03506102412939072, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.02895826380699873, + "weighted_rkd_loss": 0.01753051206469536 + }, + { + "epoch": 1.2730318257956448, + "flow_matching_loss": 0.026551509276032448, + "grad_norm": 0.23761029541492462, + "learning_rate": 2.9158015893200504e-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": 0.014330549072474242, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.007122480403631925, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.035783578641712666, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.026551509276032448, + "weighted_rkd_loss": 0.017891789320856333 + }, + { + "epoch": 1.2897822445561138, + "flow_matching_loss": 0.029302243888378144, + "grad_norm": 0.25727933645248413, + "learning_rate": 2.9124935662007137e-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": 0.014529280830174685, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0068877630867064, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.035946324840188025, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.029302243888378144, + "weighted_rkd_loss": 0.017973162420094013 + }, + { + "epoch": 1.3065326633165828, + "flow_matching_loss": 0.028624637983739376, + "grad_norm": 0.29517701268196106, + "learning_rate": 2.9091237560369467e-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": 0.016275989823043346, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.007523399125784635, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.0400753790512681, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.028624637983739376, + "weighted_rkd_loss": 0.02003768952563405 + }, + { + "epoch": 1.3232830820770518, + "flow_matching_loss": 0.028271349146962166, + "grad_norm": 0.25125834345817566, + "learning_rate": 2.905692306235163e-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": 0.01371820280328393, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.007034803554415703, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.03447120953351259, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.028271349146962166, + "weighted_rkd_loss": 0.017235604766756295 + }, + { + "epoch": 1.3400335008375208, + "flow_matching_loss": 0.02792028449475765, + "grad_norm": 0.2755165100097656, + "learning_rate": 2.9021993668980925e-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": 0.014490864146500826, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.006647658534348011, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.03562938645482063, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.02792028449475765, + "weighted_rkd_loss": 0.017814693227410316 + }, + { + "epoch": 1.3567839195979898, + "flow_matching_loss": 0.02856957260519266, + "grad_norm": 0.353760689496994, + "learning_rate": 2.8986450908182185e-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": 0.013120757881551981, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.006197075080126524, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.03243859093636274, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.02856957260519266, + "weighted_rkd_loss": 0.01621929546818137 + }, + { + "epoch": 1.3735343383584588, + "flow_matching_loss": 0.027433456107974052, + "grad_norm": 0.2630797326564789, + "learning_rate": 2.8950296334710902e-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": 0.014867633581161499, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.007008401164785028, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.036743668280541895, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.027433456107974052, + "weighted_rkd_loss": 0.018371834140270948 + }, + { + "epoch": 1.3902847571189278, + "flow_matching_loss": 0.029373966716229917, + "grad_norm": 0.25047779083251953, + "learning_rate": 2.8913531530085248e-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": 0.015035781264305114, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.007420615572482347, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.03749217819422483, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.029373966716229917, + "weighted_rkd_loss": 0.018746089097112416 + }, + { + "epoch": 1.4070351758793969, + "flow_matching_loss": 0.02783653810620308, + "grad_norm": 0.24933220446109772, + "learning_rate": 2.887615810251687e-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": 0.014838722348213196, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0071543951518833636, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.03683183994144201, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.02783653810620308, + "weighted_rkd_loss": 0.018415919970721006 + }, + { + "epoch": 1.4237855946398659, + "flow_matching_loss": 0.02553834468126297, + "grad_norm": 0.2722643315792084, + "learning_rate": 2.8838177686840577e-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": 0.01331262132152915, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0062260095495730635, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.03285125233232975, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.02553834468126297, + "weighted_rkd_loss": 0.016425626166164876 + }, + { + "epoch": 1.4405360134003349, + "flow_matching_loss": 0.02865916658192873, + "grad_norm": 0.3012257218360901, + "learning_rate": 2.8799591944442776e-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": 0.013619011919945478, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.006813362147659064, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.034051386080682276, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.02865916658192873, + "weighted_rkd_loss": 0.017025693040341138 + }, + { + "epoch": 1.457286432160804, + "flow_matching_loss": 0.02550832610577345, + "grad_norm": 0.28759023547172546, + "learning_rate": 2.8760402563188848e-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": 0.013887122552841902, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0066726688295602795, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.0344469141215086, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.02550832610577345, + "weighted_rkd_loss": 0.0172234570607543 + }, + { + "epoch": 1.474036850921273, + "flow_matching_loss": 0.029250646568834783, + "grad_norm": 0.3203338384628296, + "learning_rate": 2.8720611257349285e-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": 0.013954385695979, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.007791683613322675, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.03570045493543148, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.029250646568834783, + "weighted_rkd_loss": 0.01785022746771574 + }, + { + "epoch": 1.490787269681742, + "flow_matching_loss": 0.02804416585713625, + "grad_norm": 0.25848525762557983, + "learning_rate": 2.868021976752471e-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": 0.014760543219745159, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0070793390274047855, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.03660042583942413, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.02804416585713625, + "weighted_rkd_loss": 0.018300212919712067 + }, + { + "epoch": 1.507537688442211, + "flow_matching_loss": 0.026911542378365995, + "grad_norm": 0.28970643877983093, + "learning_rate": 2.8639229860569738e-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": 0.012177168950438499, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.006474159099161625, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.030828496813774107, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.026911542378365995, + "weighted_rkd_loss": 0.015414248406887054 + }, + { + "epoch": 1.52428810720268, + "flow_matching_loss": 0.025630979984998702, + "grad_norm": 0.22555437684059143, + "learning_rate": 2.8597643329515688e-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": 0.013054642546921968, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.006073564128018916, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.032182848826050756, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.025630979984998702, + "weighted_rkd_loss": 0.016091424413025378 + }, + { + "epoch": 1.541038525963149, + "flow_matching_loss": 0.030764632672071458, + "grad_norm": 0.2416084259748459, + "learning_rate": 2.855546199349214e-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": 0.01417373651638627, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.00662791975773871, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.03497539311647415, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.030764632672071458, + "weighted_rkd_loss": 0.017487696558237075 + }, + { + "epoch": 1.557788944723618, + "flow_matching_loss": 0.02624164503067732, + "grad_norm": 0.331834614276886, + "learning_rate": 2.85126876976474e-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": 0.013679224532097578, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.006975525338202715, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.03433397449553013, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.02624164503067732, + "weighted_rkd_loss": 0.017166987247765064 + }, + { + "epoch": 1.574539363484087, + "flow_matching_loss": 0.024381680972874166, + "grad_norm": 0.31837213039398193, + "learning_rate": 2.8469322313067726e-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": 0.013315047975629568, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.005786149669438601, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.03241624571382999, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.024381680972874166, + "weighted_rkd_loss": 0.016208122856914996 + }, + { + "epoch": 1.591289782244556, + "flow_matching_loss": 0.028025544807314874, + "grad_norm": 0.2301408052444458, + "learning_rate": 2.8425367736695522e-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": 0.012464443966746331, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.006268363306298852, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.031197250820696354, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.028025544807314874, + "weighted_rkd_loss": 0.015598625410348177 + }, + { + "epoch": 1.608040201005025, + "flow_matching_loss": 0.02433723080903292, + "grad_norm": 0.24389685690402985, + "learning_rate": 2.838082589124636e-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": 0.013607794232666493, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.006303474889136851, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.033519063331186774, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.02433723080903292, + "weighted_rkd_loss": 0.016759531665593387 + }, + { + "epoch": 1.624790619765494, + "flow_matching_loss": 0.022779523208737374, + "grad_norm": 0.28541818261146545, + "learning_rate": 2.8335698725124843e-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": 0.014848234411329032, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.006937592523172498, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.036634061485528946, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.022779523208737374, + "weighted_rkd_loss": 0.018317030742764473 + }, + { + "epoch": 1.641541038525963, + "flow_matching_loss": 0.023877214454114438, + "grad_norm": 0.2595161497592926, + "learning_rate": 2.8289988212339422e-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": 0.01390945939347148, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.006691856821998954, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.034510775655508044, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.023877214454114438, + "weighted_rkd_loss": 0.017255387827754022 + }, + { + "epoch": 1.658291457286432, + "flow_matching_loss": 0.022958588413894178, + "grad_norm": 0.27092212438583374, + "learning_rate": 2.8243696352415997e-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": 0.013638240005820989, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.006506198924034834, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.03378267884254456, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.022958588413894178, + "weighted_rkd_loss": 0.01689133942127228 + }, + { + "epoch": 1.675041876046901, + "flow_matching_loss": 0.027839783951640128, + "grad_norm": 0.2704690396785736, + "learning_rate": 2.8196825170310495e-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": 0.013532010465860366, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.006611468456685543, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.03367548901587725, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.027839783951640128, + "weighted_rkd_loss": 0.016837744507938623 + } + ], + "logging_steps": 10, + "max_steps": 5000, + "num_input_tokens_seen": 0, + "num_train_epochs": 9, + "save_steps": 5001, + "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": 64, + "trial_name": null, + "trial_params": null +} diff --git a/FewShot/39tasks/5shot/FromRKD-Timewarp_FullactionHead/seed1/default/train_5shot_seed1_fullactionhead_5k/checkpoint-3000/config.json b/FewShot/39tasks/5shot/FromRKD-Timewarp_FullactionHead/seed1/default/train_5shot_seed1_fullactionhead_5k/checkpoint-3000/config.json new file mode 100644 index 0000000000000000000000000000000000000000..4740d0455bef31c469ee5dab6d4a47d64df61665 --- /dev/null +++ b/FewShot/39tasks/5shot/FromRKD-Timewarp_FullactionHead/seed1/default/train_5shot_seed1_fullactionhead_5k/checkpoint-3000/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": 0.5, + "rkd_loss_weight_end": 0.0, + "rkd_loss_weight_schedule": null, + "rkd_loss_weight_start": 0.1, + "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/FewShot/39tasks/5shot/FromRKD-Timewarp_FullactionHead/seed1/default/train_5shot_seed1_fullactionhead_5k/checkpoint-3000/experiment_cfg/metadata.json b/FewShot/39tasks/5shot/FromRKD-Timewarp_FullactionHead/seed1/default/train_5shot_seed1_fullactionhead_5k/checkpoint-3000/experiment_cfg/metadata.json new file mode 100644 index 0000000000000000000000000000000000000000..27700cac5d292d4d42897b3c2e80bdc58e55c60e --- /dev/null +++ b/FewShot/39tasks/5shot/FromRKD-Timewarp_FullactionHead/seed1/default/train_5shot_seed1_fullactionhead_5k/checkpoint-3000/experiment_cfg/metadata.json @@ -0,0 +1,431 @@ +{ + "new_embodiment": { + "statistics": { + "state": { + "base_position": { + "max": [ + 7.778976917266846, + 0.33403685688972473, + 0.7223060727119446 + ], + "min": [ + -4.821934223175049, + -6.785110950469971, + 0.6994727253913879 + ], + "mean": [ + 2.5719731235192254, + -2.1489371002267235, + 0.7003612741268546 + ], + "std": [ + 1.669945884010784, + 1.3739411381361726, + 0.0009413282367754987 + ], + "q01": [ + -0.7944286339399775, + -5.486611518471242, + 0.699977555971739 + ], + "q99": [ + 6.019886905263648, + -0.5229184210026855, + 0.702930326282967 + ] + }, + "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, + 0.248543484335273, + 0.5959317405459531 + ], + "std": [ + 0.0, + 0.0, + 0.6655550240074092, + 0.374283995143559 + ], + "q01": [ + 0.0, + 0.0, + -1.0000000000000002, + 4.503254968910557e-06 + ], + "q99": [ + 0.0, + 0.0, + 1.0000000000000002, + 1.0000000000000002 + ] + }, + "end_effector_position_relative": { + "max": [ + 0.8848381638526917, + 0.8569459915161133, + 1.053884744644165 + ], + "min": [ + -0.5828549861907959, + -0.8486236333847046, + -0.3446122705936432 + ], + "mean": [ + 0.2465490874130862, + -0.004711658908481467, + 0.43463835464312073 + ], + "std": [ + 0.15798784670671595, + 0.25624102400497073, + 0.20836028727974915 + ], + "q01": [ + -0.020752051134640608, + -0.4837933887347875, + 0.1802942952289915 + ], + "q99": [ + 0.5386784714844007, + 0.4801424212024721, + 0.695095732975766 + ] + }, + "end_effector_rotation_relative": { + "max": [ + 0.9999949932098389, + 0.9997354745864868, + 0.9900747537612915, + 0.9735639095306396 + ], + "min": [ + -0.9999914169311523, + -0.9999884366989136, + -0.9975369572639465, + 6.754255821306288e-08 + ], + "mean": [ + -0.22531859714056926, + 0.046973702821026614, + -0.0568176967878771, + 0.12364307042955697 + ], + "std": [ + 0.8502144333612052, + 0.29591250318147405, + 0.30218053156837016, + 0.1636031353412931 + ], + "q01": [ + -0.9959057387172566, + -0.5665384745573048, + -0.5329334620834977, + 0.0014372279999397553 + ], + "q99": [ + 0.9904107951565839, + 0.6359623612021396, + 0.5026076022439377, + 0.42955559234879914 + ] + }, + "gripper_qpos": { + "max": [ + 0.05289580300450325, + 0.00937709677964449 + ], + "min": [ + -0.011669117026031017, + -0.06307756155729294 + ], + "mean": [ + 0.03188757060429838, + -0.0321793333086053 + ], + "std": [ + 0.01171052313737148, + 0.011427336801203189 + ], + "q01": [ + 0.006462901319459925, + -0.040557140599419166 + ], + "q99": [ + 0.04037619335388386, + -0.006523947835993347 + ] + } + }, + "action": { + "base_motion": { + "max": [ + 1.0, + 1.0, + 1.0, + 0.0 + ], + "min": [ + -1.0, + -1.0, + -1.0, + 0.0 + ], + "mean": [ + 0.006639007355857881, + 0.0027330421795561874, + -0.0017774441970214892, + 0.0 + ], + "std": [ + 0.11841528114040736, + 0.1041359083946544, + 0.08346265521208351, + 0.0 + ], + "q01": [ + -0.17562359675270237, + -0.21706402321529558, + -0.16273392585723104, + 0.0 + ], + "q99": [ + 0.2572918701644887, + 0.27022325575787115, + 0.1369626751049278, + 0.0 + ] + }, + "control_mode": { + "max": [ + 1.0 + ], + "min": [ + -1.0 + ], + "mean": [ + -0.9205539608927732 + ], + "std": [ + 0.39063657242534666 + ], + "q01": [ + -1.0 + ], + "q99": [ + -0.17711863866562935 + ] + }, + "end_effector_position": { + "max": [ + 1.0, + 1.0, + 1.0 + ], + "min": [ + -1.0, + -1.0, + -1.0 + ], + "mean": [ + -0.00924803964621179, + 0.004053841761390575, + -0.08385115578460103 + ], + "std": [ + 0.45413072518223824, + 0.40158612714634856, + 0.41068634292849304 + ], + "q01": [ + -0.9669817145875343, + -0.9241307799783718, + -0.8264079503068825 + ], + "q99": [ + 0.8021511975722166, + 0.9241346582748478, + 0.8348116781502298 + ] + }, + "end_effector_rotation": { + "max": [ + 1.0, + 1.0, + 1.0 + ], + "min": [ + -1.0, + -1.0, + -1.0 + ], + "mean": [ + 0.007644157504353809, + -0.019962543093886335, + -0.002047110452602617 + ], + "std": [ + 0.11076012154791087, + 0.13176735140285795, + 0.12286528665895614 + ], + "q01": [ + -0.2659105439401192, + -0.3961694033813533, + -0.3434870832468936 + ], + "q99": [ + 0.3145786959266369, + 0.31309430361227547, + 0.34107754401942986 + ] + }, + "gripper_close": { + "max": [ + 1.0 + ], + "min": [ + -1.0 + ], + "mean": [ + -0.29870612267533575 + ], + "std": [ + 0.9543487555375691 + ], + "q01": [ + -1.0 + ], + "q99": [ + 0.7610674802484976 + ] + } + } + }, + "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/FewShot/39tasks/5shot/FromRKD-Timewarp_FullactionHead/seed1/default/train_5shot_seed1_fullactionhead_5k/checkpoint-3000/model.safetensors.index.json b/FewShot/39tasks/5shot/FromRKD-Timewarp_FullactionHead/seed1/default/train_5shot_seed1_fullactionhead_5k/checkpoint-3000/model.safetensors.index.json new file mode 100644 index 0000000000000000000000000000000000000000..b20baed71f4d16a411223b70d45d934c6cbbfe97 --- /dev/null +++ b/FewShot/39tasks/5shot/FromRKD-Timewarp_FullactionHead/seed1/default/train_5shot_seed1_fullactionhead_5k/checkpoint-3000/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/FewShot/39tasks/5shot/FromRKD-Timewarp_FullactionHead/seed1/default/train_5shot_seed1_fullactionhead_5k/checkpoint-3000/trainer_state.json b/FewShot/39tasks/5shot/FromRKD-Timewarp_FullactionHead/seed1/default/train_5shot_seed1_fullactionhead_5k/checkpoint-3000/trainer_state.json new file mode 100644 index 0000000000000000000000000000000000000000..c42c660448d2144d998d33f55772d9d44195f18a --- /dev/null +++ b/FewShot/39tasks/5shot/FromRKD-Timewarp_FullactionHead/seed1/default/train_5shot_seed1_fullactionhead_5k/checkpoint-3000/trainer_state.json @@ -0,0 +1,13834 @@ +{ + "best_global_step": null, + "best_metric": null, + "best_model_checkpoint": null, + "epoch": 5.025125628140704, + "eval_steps": 500, + "global_step": 3000, + "is_hyper_param_search": false, + "is_local_process_zero": true, + "is_world_process_zero": true, + "log_history": [ + { + "epoch": 0.01675041876046901, + "flow_matching_loss": 0.18734376579523088, + "grad_norm": 1.2580132484436035, + "learning_rate": 1.08e-06, + "loss": 0.384, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.13968379348516463, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.11389333754777908, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.3932609289884567, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.18734376579523088, + "weighted_rkd_loss": 0.19663046449422836 + }, + { + "epoch": 0.03350083752093802, + "flow_matching_loss": 0.19343584030866623, + "grad_norm": 0.9545837044715881, + "learning_rate": 2.28e-06, + "loss": 0.3738, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.12992414832115173, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.10088557079434395, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.3607338696718216, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.19343584030866623, + "weighted_rkd_loss": 0.1803669348359108 + }, + { + "epoch": 0.05025125628140704, + "flow_matching_loss": 0.16988817751407623, + "grad_norm": 0.9067039489746094, + "learning_rate": 3.48e-06, + "loss": 0.3561, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.13251580074429511, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.1073499672114849, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.37238157391548155, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.16988817751407623, + "weighted_rkd_loss": 0.18619078695774077 + }, + { + "epoch": 0.06700167504187604, + "flow_matching_loss": 0.1650318592786789, + "grad_norm": 1.1154892444610596, + "learning_rate": 4.68e-06, + "loss": 0.3387, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.12463015466928482, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.09805893376469613, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.34731924533843994, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.1650318592786789, + "weighted_rkd_loss": 0.17365962266921997 + }, + { + "epoch": 0.08375209380234507, + "flow_matching_loss": 0.13135033324360848, + "grad_norm": 0.7893933653831482, + "learning_rate": 5.8800000000000005e-06, + "loss": 0.2851, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.11209887266159058, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.08326855897903443, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.30746630430221555, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.13135033324360848, + "weighted_rkd_loss": 0.15373315215110778 + }, + { + "epoch": 0.10050251256281408, + "flow_matching_loss": 0.11830725595355034, + "grad_norm": 0.6967894434928894, + "learning_rate": 7.08e-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": 0.1122154638171196, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.08148682340979577, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.30591775476932526, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.11830725595355034, + "weighted_rkd_loss": 0.15295887738466263 + }, + { + "epoch": 0.11725293132328309, + "flow_matching_loss": 0.09991217777132988, + "grad_norm": 0.5718310475349426, + "learning_rate": 8.28e-06, + "loss": 0.2493, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.11076194941997528, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.07722590751945972, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.29874980747699736, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.09991217777132988, + "weighted_rkd_loss": 0.14937490373849868 + }, + { + "epoch": 0.13400335008375208, + "flow_matching_loss": 0.09349806271493435, + "grad_norm": 0.5672374367713928, + "learning_rate": 9.48e-06, + "loss": 0.232, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.10319361835718155, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.07060621418058873, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.2769934505224228, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.09349806271493435, + "weighted_rkd_loss": 0.1384967252612114 + }, + { + "epoch": 0.1507537688442211, + "flow_matching_loss": 0.08115713745355606, + "grad_norm": 0.5059905052185059, + "learning_rate": 1.068e-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": 0.09615755379199982, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0642927698791027, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.2566078782081604, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.08115713745355606, + "weighted_rkd_loss": 0.1283039391040802 + }, + { + "epoch": 0.16750418760469013, + "flow_matching_loss": 0.07341464199125766, + "grad_norm": 0.5954922437667847, + "learning_rate": 1.1880000000000001e-05, + "loss": 0.1941, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.09074362516403198, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.05983745716512203, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.24132470935583114, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.07341464199125766, + "weighted_rkd_loss": 0.12066235467791557 + }, + { + "epoch": 0.18425460636515914, + "flow_matching_loss": 0.06752067059278488, + "grad_norm": 0.53483647108078, + "learning_rate": 1.308e-05, + "loss": 0.1833, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.08770871832966805, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0561994768679142, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.23161691278219224, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.06752067059278488, + "weighted_rkd_loss": 0.11580845639109612 + }, + { + "epoch": 0.20100502512562815, + "flow_matching_loss": 0.0645824696868658, + "grad_norm": 0.4954916536808014, + "learning_rate": 1.428e-05, + "loss": 0.1694, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.08025030866265297, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.049181188270449636, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.20968180298805236, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.0645824696868658, + "weighted_rkd_loss": 0.10484090149402618 + }, + { + "epoch": 0.21775544388609716, + "flow_matching_loss": 0.06256739944219589, + "grad_norm": 0.6052863001823425, + "learning_rate": 1.548e-05, + "loss": 0.1603, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.07429649494588375, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.04693063050508499, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.19552361965179443, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.06256739944219589, + "weighted_rkd_loss": 0.09776180982589722 + }, + { + "epoch": 0.23450586264656617, + "flow_matching_loss": 0.0674729149788618, + "grad_norm": 0.6622745990753174, + "learning_rate": 1.6680000000000003e-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": 0.07380652725696564, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.04217908121645451, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.18979213535785674, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.0674729149788618, + "weighted_rkd_loss": 0.09489606767892837 + }, + { + "epoch": 0.25125628140703515, + "flow_matching_loss": 0.05902594663202763, + "grad_norm": 0.596757173538208, + "learning_rate": 1.7879999999999998e-05, + "loss": 0.1449, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.06658526808023453, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.03863737247884273, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.1718079075217247, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.05902594663202763, + "weighted_rkd_loss": 0.08590395376086235 + }, + { + "epoch": 0.26800670016750416, + "flow_matching_loss": 0.062078724056482314, + "grad_norm": 0.5115512609481812, + "learning_rate": 1.908e-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": 0.06346513852477073, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.03915274403989315, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.1660830244421959, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.062078724056482314, + "weighted_rkd_loss": 0.08304151222109794 + }, + { + "epoch": 0.2847571189279732, + "flow_matching_loss": 0.05774013251066208, + "grad_norm": 0.4801546335220337, + "learning_rate": 2.0280000000000002e-05, + "loss": 0.1378, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.061606602370738985, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.03699663132429123, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.1602098360657692, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.05774013251066208, + "weighted_rkd_loss": 0.0801049180328846 + }, + { + "epoch": 0.3015075376884422, + "flow_matching_loss": 0.05709497965872288, + "grad_norm": 0.41434067487716675, + "learning_rate": 2.148e-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": 0.0580575879663229, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0330570163205266, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.1491721898317337, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.05709497965872288, + "weighted_rkd_loss": 0.07458609491586685 + }, + { + "epoch": 0.31825795644891125, + "flow_matching_loss": 0.05621899552643299, + "grad_norm": 0.4601005017757416, + "learning_rate": 2.268e-05, + "loss": 0.1284, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.056104433909058574, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.03210239075124264, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.1443112574517727, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.05621899552643299, + "weighted_rkd_loss": 0.07215562872588635 + }, + { + "epoch": 0.33500837520938026, + "flow_matching_loss": 0.057572386041283605, + "grad_norm": 0.5096933245658875, + "learning_rate": 2.3880000000000002e-05, + "loss": 0.129, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.05570494718849659, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.03144418653100729, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.1428540788590908, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.057572386041283605, + "weighted_rkd_loss": 0.0714270394295454 + }, + { + "epoch": 0.35175879396984927, + "flow_matching_loss": 0.052765559032559396, + "grad_norm": 0.4708020091056824, + "learning_rate": 2.508e-05, + "loss": 0.1177, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.050074973702430726, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.029708896577358247, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.1298588439822197, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.052765559032559396, + "weighted_rkd_loss": 0.06492942199110985 + }, + { + "epoch": 0.3685092127303183, + "flow_matching_loss": 0.048277949541807176, + "grad_norm": 0.41520747542381287, + "learning_rate": 2.628e-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": 0.0443026889115572, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.024513646587729455, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.11311902478337288, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.048277949541807176, + "weighted_rkd_loss": 0.05655951239168644 + }, + { + "epoch": 0.3852596314907873, + "flow_matching_loss": 0.050647230073809626, + "grad_norm": 0.488421767950058, + "learning_rate": 2.748e-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": 0.044228341057896615, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.02473187427967787, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.11318855583667756, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.050647230073809626, + "weighted_rkd_loss": 0.05659427791833878 + }, + { + "epoch": 0.4020100502512563, + "flow_matching_loss": 0.052835336327552794, + "grad_norm": 0.4683481752872467, + "learning_rate": 2.868e-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": 0.043489795550704, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.025273922458291055, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.11225351244211197, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.052835336327552794, + "weighted_rkd_loss": 0.056126756221055986 + }, + { + "epoch": 0.4187604690117253, + "flow_matching_loss": 0.0462393905967474, + "grad_norm": 0.3933083415031433, + "learning_rate": 2.9880000000000002e-05, + "loss": 0.0989, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.04124091118574143, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.022835531365126373, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.10531735345721245, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.0462393905967474, + "weighted_rkd_loss": 0.052658676728606225 + }, + { + "epoch": 0.4355108877721943, + "flow_matching_loss": 0.04874549731612206, + "grad_norm": 0.40502575039863586, + "learning_rate": 2.9999734259635676e-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": 0.039895395189523696, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.021268937550485135, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.10105972811579704, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.04874549731612206, + "weighted_rkd_loss": 0.05052986405789852 + }, + { + "epoch": 0.45226130653266333, + "flow_matching_loss": 0.051396344974637034, + "grad_norm": 0.5639358162879944, + "learning_rate": 2.9998815663057245e-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": 0.03602162189781666, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.020294071175158024, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.09233731552958488, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.051396344974637034, + "weighted_rkd_loss": 0.04616865776479244 + }, + { + "epoch": 0.46901172529313234, + "flow_matching_loss": 0.04447287768125534, + "grad_norm": 0.40996530652046204, + "learning_rate": 2.999724096969228e-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": 0.0336500596255064, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.017429837398231028, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.08472995683550835, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.04447287768125534, + "weighted_rkd_loss": 0.042364978417754176 + }, + { + "epoch": 0.48576214405360135, + "flow_matching_loss": 0.04591554999351501, + "grad_norm": 0.4135409891605377, + "learning_rate": 2.9995010248422984e-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": 0.034760774858295915, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.018106953613460063, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.08762850314378738, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.04591554999351501, + "weighted_rkd_loss": 0.04381425157189369 + }, + { + "epoch": 0.5025125628140703, + "flow_matching_loss": 0.04205652140080929, + "grad_norm": 0.38845568895339966, + "learning_rate": 2.999212359682832e-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": 0.03268313799053431, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.017657354567199947, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.08302363082766533, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.04205652140080929, + "weighted_rkd_loss": 0.04151181541383266 + }, + { + "epoch": 0.5192629815745393, + "flow_matching_loss": 0.04226362742483616, + "grad_norm": 0.3450797200202942, + "learning_rate": 2.9988581141179787e-05, + "loss": 0.0805, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.030486975982785224, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.015418309066444636, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.07639226131141186, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.04226362742483616, + "weighted_rkd_loss": 0.03819613065570593 + }, + { + "epoch": 0.5360134003350083, + "flow_matching_loss": 0.0407985620200634, + "grad_norm": 0.3774641752243042, + "learning_rate": 2.998438303643587e-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": 0.03008116353303194, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.01613120548427105, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.07629353329539298, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.0407985620200634, + "weighted_rkd_loss": 0.03814676664769649 + }, + { + "epoch": 0.5527638190954773, + "flow_matching_loss": 0.041433697938919066, + "grad_norm": 0.37684324383735657, + "learning_rate": 2.997952946623527e-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": 0.03052981775254011, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.015478260163217784, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.07653789557516574, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.041433697938919066, + "weighted_rkd_loss": 0.03826894778758287 + }, + { + "epoch": 0.5695142378559463, + "flow_matching_loss": 0.04276835490018129, + "grad_norm": 0.41635242104530334, + "learning_rate": 2.9974020642888875e-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": 0.02872589062899351, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.014681840222328902, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.07213362231850624, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.04276835490018129, + "weighted_rkd_loss": 0.03606681115925312 + }, + { + "epoch": 0.5862646566164154, + "flow_matching_loss": 0.04388797134160995, + "grad_norm": 0.4484618604183197, + "learning_rate": 2.9967856807370458e-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": 0.03199950344860554, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.01809648834168911, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.08209549561142922, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.04388797134160995, + "weighted_rkd_loss": 0.04104774780571461 + }, + { + "epoch": 0.6030150753768844, + "flow_matching_loss": 0.04277344048023224, + "grad_norm": 0.4147380590438843, + "learning_rate": 2.996103822930615e-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": 0.029685858823359012, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.015408230014145374, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.07477994747459889, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.04277344048023224, + "weighted_rkd_loss": 0.037389973737299444 + }, + { + "epoch": 0.6197654941373534, + "flow_matching_loss": 0.04038205854594708, + "grad_norm": 0.371023952960968, + "learning_rate": 2.9953565206962653e-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": 0.028526164032518862, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.01493225758895278, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.07198458574712277, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.04038205854594708, + "weighted_rkd_loss": 0.035992292873561385 + }, + { + "epoch": 0.6365159128978225, + "flow_matching_loss": 0.04215538240969181, + "grad_norm": 0.3584204316139221, + "learning_rate": 2.9945438067234172e-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": 0.027064095251262188, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.014362859819084407, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.06849104985594749, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.04215538240969181, + "weighted_rkd_loss": 0.034245524927973746 + }, + { + "epoch": 0.6532663316582915, + "flow_matching_loss": 0.03905519917607307, + "grad_norm": 0.3805311620235443, + "learning_rate": 2.9936657165628124e-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": 0.026047029718756675, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.013608651608228684, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.06570271104574203, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.03905519917607307, + "weighted_rkd_loss": 0.032851355522871016 + }, + { + "epoch": 0.6700167504187605, + "flow_matching_loss": 0.03584225699305534, + "grad_norm": 0.35377806425094604, + "learning_rate": 2.992722288624959e-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": 0.02628317419439554, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.013779552932828665, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.06634590104222297, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.03584225699305534, + "weighted_rkd_loss": 0.03317295052111149 + }, + { + "epoch": 0.6867671691792295, + "flow_matching_loss": 0.03798154555261135, + "grad_norm": 0.33630454540252686, + "learning_rate": 2.9917135641784515e-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": 0.025117580965161323, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.013396832998842001, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.06363199502229691, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.03798154555261135, + "weighted_rkd_loss": 0.031815997511148456 + }, + { + "epoch": 0.7035175879396985, + "flow_matching_loss": 0.03501146622002125, + "grad_norm": 0.3093757629394531, + "learning_rate": 2.9906395873481655e-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": 0.024045810103416443, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.012114135175943374, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.06020575501024723, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.03501146622002125, + "weighted_rkd_loss": 0.030102877505123617 + }, + { + "epoch": 0.7202680067001676, + "flow_matching_loss": 0.038727116957306865, + "grad_norm": 0.2813211977481842, + "learning_rate": 2.989500405113326e-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": 0.025140311382710934, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.01257405923679471, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.0628546815365553, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.038727116957306865, + "weighted_rkd_loss": 0.03142734076827765 + }, + { + "epoch": 0.7370184254606366, + "flow_matching_loss": 0.035598259046673776, + "grad_norm": 0.3280915319919586, + "learning_rate": 2.9882960673054547e-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": 0.02384033761918545, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.012014199746772647, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.059694874286651614, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.035598259046673776, + "weighted_rkd_loss": 0.029847437143325807 + }, + { + "epoch": 0.7537688442211056, + "flow_matching_loss": 0.03603664711117745, + "grad_norm": 0.30604344606399536, + "learning_rate": 2.9870266266061886e-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": 0.022386168129742146, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.01189499134197831, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.05666732750833035, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.03603664711117745, + "weighted_rkd_loss": 0.028333663754165173 + }, + { + "epoch": 0.7705192629815746, + "flow_matching_loss": 0.036646503023803234, + "grad_norm": 0.3464244604110718, + "learning_rate": 2.9856921385449766e-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": 0.02474432121962309, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.012881245650351047, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.06236988790333271, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.036646503023803234, + "weighted_rkd_loss": 0.031184943951666357 + }, + { + "epoch": 0.7872696817420436, + "flow_matching_loss": 0.03805769421160221, + "grad_norm": 0.34830617904663086, + "learning_rate": 2.9842926614966502e-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": 0.02307860888540745, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.01193495150655508, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.05809217020869255, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.03805769421160221, + "weighted_rkd_loss": 0.029046085104346274 + }, + { + "epoch": 0.8040201005025126, + "flow_matching_loss": 0.03244946617633104, + "grad_norm": 0.36722618341445923, + "learning_rate": 2.982828256678868e-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": 0.02207295410335064, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.010983642470091582, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.05512955039739609, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.03244946617633104, + "weighted_rkd_loss": 0.027564775198698044 + }, + { + "epoch": 0.8207705192629816, + "flow_matching_loss": 0.035172288119792935, + "grad_norm": 0.2739450931549072, + "learning_rate": 2.9812989881494412e-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": 0.02087551765143871, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.010237869061529636, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.05198890343308449, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.035172288119792935, + "weighted_rkd_loss": 0.025994451716542244 + }, + { + "epoch": 0.8375209380234506, + "flow_matching_loss": 0.03737096916884184, + "grad_norm": 0.32612043619155884, + "learning_rate": 2.9797049228035296e-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": 0.02040377100929618, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.009668528474867344, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.05047607012093067, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.03737096916884184, + "weighted_rkd_loss": 0.025238035060465337 + }, + { + "epoch": 0.8542713567839196, + "flow_matching_loss": 0.03377289455384016, + "grad_norm": 0.29827985167503357, + "learning_rate": 2.978046130370715e-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": 0.019000360555946828, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.008991975570097566, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.04699269644916058, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.03377289455384016, + "weighted_rkd_loss": 0.02349634822458029 + }, + { + "epoch": 0.8710217755443886, + "flow_matching_loss": 0.034123093634843824, + "grad_norm": 0.2898612320423126, + "learning_rate": 2.9763226834119535e-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": 0.020308055356144904, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.010839158948510886, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.05145526938140392, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.034123093634843824, + "weighted_rkd_loss": 0.02572763469070196 + }, + { + "epoch": 0.8877721943048577, + "flow_matching_loss": 0.0334856515750289, + "grad_norm": 0.30531373620033264, + "learning_rate": 2.9745346573163963e-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": 0.022741097211837768, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.01176437777467072, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.05724657289683819, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.0334856515750289, + "weighted_rkd_loss": 0.028623286448419095 + }, + { + "epoch": 0.9045226130653267, + "flow_matching_loss": 0.03164167143404484, + "grad_norm": 0.38043874502182007, + "learning_rate": 2.9726821302980987e-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": 0.02146173259243369, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.010472412640228867, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.05339587815105915, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.03164167143404484, + "weighted_rkd_loss": 0.026697939075529576 + }, + { + "epoch": 0.9212730318257957, + "flow_matching_loss": 0.03139312732964754, + "grad_norm": 0.358161985874176, + "learning_rate": 2.9707651833925935e-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": 0.018665229622274637, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.00916362195275724, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.046494081243872644, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.03139312732964754, + "weighted_rkd_loss": 0.023247040621936322 + }, + { + "epoch": 0.9380234505862647, + "flow_matching_loss": 0.03471959326416254, + "grad_norm": 0.383441299200058, + "learning_rate": 2.968783900453348e-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": 0.02090346897020936, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.010127922613173723, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.05193486027419567, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.03471959326416254, + "weighted_rkd_loss": 0.025967430137097836 + }, + { + "epoch": 0.9547738693467337, + "flow_matching_loss": 0.03342412430793047, + "grad_norm": 0.3316456973552704, + "learning_rate": 2.9667383681480968e-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": 0.018697252310812474, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.00936074941419065, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.04675525352358818, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.03342412430793047, + "weighted_rkd_loss": 0.02337762676179409 + }, + { + "epoch": 0.9715242881072027, + "flow_matching_loss": 0.034992733597755434, + "grad_norm": 0.31459757685661316, + "learning_rate": 2.9646286759550493e-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": 0.019015263859182598, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.009189907228574157, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.04722043462097645, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.034992733597755434, + "weighted_rkd_loss": 0.023610217310488225 + }, + { + "epoch": 0.9882747068676717, + "flow_matching_loss": 0.03335383273661137, + "grad_norm": 0.39081528782844543, + "learning_rate": 2.9624549161589764e-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": 0.0166207866743207, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.007937656342983246, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.041179229691624644, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.03335383273661137, + "weighted_rkd_loss": 0.020589614845812322 + }, + { + "epoch": 1.0050251256281406, + "flow_matching_loss": 0.030267693661153318, + "grad_norm": 0.33321326971054077, + "learning_rate": 2.9602171838471745e-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": 0.02047300133854151, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.009730400750413537, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.05067640393972397, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.030267693661153318, + "weighted_rkd_loss": 0.025338201969861984 + }, + { + "epoch": 1.0217755443886096, + "flow_matching_loss": 0.028719464503228664, + "grad_norm": 0.33434903621673584, + "learning_rate": 2.957915576905303e-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": 0.01841264832764864, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.010385369556024671, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.047210666351020336, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.028719464503228664, + "weighted_rkd_loss": 0.023605333175510168 + }, + { + "epoch": 1.0385259631490786, + "flow_matching_loss": 0.029228216037154198, + "grad_norm": 0.31092560291290283, + "learning_rate": 2.9555501960131058e-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": 0.01791312750428915, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.008790831873193383, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.04461708702147007, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.029228216037154198, + "weighted_rkd_loss": 0.022308543510735034 + }, + { + "epoch": 1.0552763819095476, + "flow_matching_loss": 0.02980944626033306, + "grad_norm": 0.2826099097728729, + "learning_rate": 2.9531211446400068e-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": 0.018980990443378688, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.009706261800602078, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.04766824245452881, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.02980944626033306, + "weighted_rkd_loss": 0.023834121227264405 + }, + { + "epoch": 1.0720268006700167, + "flow_matching_loss": 0.030087723582983016, + "grad_norm": 0.3223755359649658, + "learning_rate": 2.9506285290405813e-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": 0.01794863734394312, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.00857453583739698, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.044471810758113864, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.030087723582983016, + "weighted_rkd_loss": 0.022235905379056932 + }, + { + "epoch": 1.0887772194304857, + "flow_matching_loss": 0.031088070012629032, + "grad_norm": 0.3262973725795746, + "learning_rate": 2.948072458249911e-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": 0.018507144786417483, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.008912034891545772, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.04592632427811623, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.031088070012629032, + "weighted_rkd_loss": 0.022963162139058114 + }, + { + "epoch": 1.1055276381909547, + "flow_matching_loss": 0.028495705127716063, + "grad_norm": 0.33219948410987854, + "learning_rate": 2.9454530440788125e-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": 0.02023590700700879, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.009690370690077544, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.0501621849834919, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.028495705127716063, + "weighted_rkd_loss": 0.02508109249174595 + }, + { + "epoch": 1.1222780569514237, + "flow_matching_loss": 0.03310795053839684, + "grad_norm": 0.3126451373100281, + "learning_rate": 2.9427704011089474e-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": 0.018809593096375466, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.008968799049034715, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.046587985008955, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.03310795053839684, + "weighted_rkd_loss": 0.0232939925044775 + }, + { + "epoch": 1.1390284757118927, + "flow_matching_loss": 0.03170393034815788, + "grad_norm": 0.275397390127182, + "learning_rate": 2.94002464668781e-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": 0.01641182703897357, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0077222629450261595, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.04054591674357653, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.03170393034815788, + "weighted_rkd_loss": 0.020272958371788264 + }, + { + "epoch": 1.1557788944723617, + "flow_matching_loss": 0.031644315272569654, + "grad_norm": 0.25714337825775146, + "learning_rate": 2.9372159009235936e-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": 0.017135247867554427, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.008725387183949352, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.04299588277935982, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.031644315272569654, + "weighted_rkd_loss": 0.02149794138967991 + }, + { + "epoch": 1.1725293132328307, + "flow_matching_loss": 0.03273461870849133, + "grad_norm": 0.22671352326869965, + "learning_rate": 2.9343442866799365e-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": 0.014938866905868053, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.006849234201945364, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.03672696817666292, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.03273461870849133, + "weighted_rkd_loss": 0.01836348408833146 + }, + { + "epoch": 1.1892797319932997, + "flow_matching_loss": 0.02922193594276905, + "grad_norm": 0.2870095670223236, + "learning_rate": 2.9314099295705495e-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": 0.016047394555062055, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.007527278270572424, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.039622067287564275, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.02922193594276905, + "weighted_rkd_loss": 0.019811033643782137 + }, + { + "epoch": 1.2060301507537687, + "flow_matching_loss": 0.026027432456612588, + "grad_norm": 0.22585110366344452, + "learning_rate": 2.928412957953718e-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": 0.014577559195458889, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.007318647485226393, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.03647376541048288, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.026027432456612588, + "weighted_rkd_loss": 0.01823688270524144 + }, + { + "epoch": 1.2227805695142377, + "flow_matching_loss": 0.029601436294615268, + "grad_norm": 0.337597131729126, + "learning_rate": 2.925353502926689e-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": 0.01586325457319617, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.007780421571806073, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.03950693104416132, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.029601436294615268, + "weighted_rkd_loss": 0.01975346552208066 + }, + { + "epoch": 1.2395309882747068, + "flow_matching_loss": 0.024688402004539967, + "grad_norm": 0.32554107904434204, + "learning_rate": 2.922231698319937e-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": 0.014900147635489702, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.006712513882666827, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.03651280924677849, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.024688402004539967, + "weighted_rkd_loss": 0.018256404623389243 + }, + { + "epoch": 1.2562814070351758, + "flow_matching_loss": 0.02895826380699873, + "grad_norm": 0.2865455746650696, + "learning_rate": 2.9190476806913088e-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": 0.014406697638332844, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.006247629038989544, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.03506102412939072, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.02895826380699873, + "weighted_rkd_loss": 0.01753051206469536 + }, + { + "epoch": 1.2730318257956448, + "flow_matching_loss": 0.026551509276032448, + "grad_norm": 0.23761029541492462, + "learning_rate": 2.9158015893200504e-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": 0.014330549072474242, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.007122480403631925, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.035783578641712666, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.026551509276032448, + "weighted_rkd_loss": 0.017891789320856333 + }, + { + "epoch": 1.2897822445561138, + "flow_matching_loss": 0.029302243888378144, + "grad_norm": 0.25727933645248413, + "learning_rate": 2.9124935662007137e-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": 0.014529280830174685, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0068877630867064, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.035946324840188025, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.029302243888378144, + "weighted_rkd_loss": 0.017973162420094013 + }, + { + "epoch": 1.3065326633165828, + "flow_matching_loss": 0.028624637983739376, + "grad_norm": 0.29517701268196106, + "learning_rate": 2.9091237560369467e-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": 0.016275989823043346, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.007523399125784635, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.0400753790512681, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.028624637983739376, + "weighted_rkd_loss": 0.02003768952563405 + }, + { + "epoch": 1.3232830820770518, + "flow_matching_loss": 0.028271349146962166, + "grad_norm": 0.25125834345817566, + "learning_rate": 2.905692306235163e-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": 0.01371820280328393, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.007034803554415703, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.03447120953351259, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.028271349146962166, + "weighted_rkd_loss": 0.017235604766756295 + }, + { + "epoch": 1.3400335008375208, + "flow_matching_loss": 0.02792028449475765, + "grad_norm": 0.2755165100097656, + "learning_rate": 2.9021993668980925e-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": 0.014490864146500826, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.006647658534348011, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.03562938645482063, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.02792028449475765, + "weighted_rkd_loss": 0.017814693227410316 + }, + { + "epoch": 1.3567839195979898, + "flow_matching_loss": 0.02856957260519266, + "grad_norm": 0.353760689496994, + "learning_rate": 2.8986450908182185e-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": 0.013120757881551981, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.006197075080126524, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.03243859093636274, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.02856957260519266, + "weighted_rkd_loss": 0.01621929546818137 + }, + { + "epoch": 1.3735343383584588, + "flow_matching_loss": 0.027433456107974052, + "grad_norm": 0.2630797326564789, + "learning_rate": 2.8950296334710902e-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": 0.014867633581161499, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.007008401164785028, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.036743668280541895, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.027433456107974052, + "weighted_rkd_loss": 0.018371834140270948 + }, + { + "epoch": 1.3902847571189278, + "flow_matching_loss": 0.029373966716229917, + "grad_norm": 0.25047779083251953, + "learning_rate": 2.8913531530085248e-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": 0.015035781264305114, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.007420615572482347, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.03749217819422483, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.029373966716229917, + "weighted_rkd_loss": 0.018746089097112416 + }, + { + "epoch": 1.4070351758793969, + "flow_matching_loss": 0.02783653810620308, + "grad_norm": 0.24933220446109772, + "learning_rate": 2.887615810251687e-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": 0.014838722348213196, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0071543951518833636, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.03683183994144201, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.02783653810620308, + "weighted_rkd_loss": 0.018415919970721006 + }, + { + "epoch": 1.4237855946398659, + "flow_matching_loss": 0.02553834468126297, + "grad_norm": 0.2722643315792084, + "learning_rate": 2.8838177686840577e-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": 0.01331262132152915, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0062260095495730635, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.03285125233232975, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.02553834468126297, + "weighted_rkd_loss": 0.016425626166164876 + }, + { + "epoch": 1.4405360134003349, + "flow_matching_loss": 0.02865916658192873, + "grad_norm": 0.3012257218360901, + "learning_rate": 2.8799591944442776e-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": 0.013619011919945478, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.006813362147659064, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.034051386080682276, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.02865916658192873, + "weighted_rkd_loss": 0.017025693040341138 + }, + { + "epoch": 1.457286432160804, + "flow_matching_loss": 0.02550832610577345, + "grad_norm": 0.28759023547172546, + "learning_rate": 2.8760402563188848e-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": 0.013887122552841902, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0066726688295602795, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.0344469141215086, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.02550832610577345, + "weighted_rkd_loss": 0.0172234570607543 + }, + { + "epoch": 1.474036850921273, + "flow_matching_loss": 0.029250646568834783, + "grad_norm": 0.3203338384628296, + "learning_rate": 2.8720611257349285e-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": 0.013954385695979, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.007791683613322675, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.03570045493543148, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.029250646568834783, + "weighted_rkd_loss": 0.01785022746771574 + }, + { + "epoch": 1.490787269681742, + "flow_matching_loss": 0.02804416585713625, + "grad_norm": 0.25848525762557983, + "learning_rate": 2.868021976752471e-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": 0.014760543219745159, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0070793390274047855, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.03660042583942413, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.02804416585713625, + "weighted_rkd_loss": 0.018300212919712067 + }, + { + "epoch": 1.507537688442211, + "flow_matching_loss": 0.026911542378365995, + "grad_norm": 0.28970643877983093, + "learning_rate": 2.8639229860569738e-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": 0.012177168950438499, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.006474159099161625, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.030828496813774107, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.026911542378365995, + "weighted_rkd_loss": 0.015414248406887054 + }, + { + "epoch": 1.52428810720268, + "flow_matching_loss": 0.025630979984998702, + "grad_norm": 0.22555437684059143, + "learning_rate": 2.8597643329515688e-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": 0.013054642546921968, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.006073564128018916, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.032182848826050756, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.025630979984998702, + "weighted_rkd_loss": 0.016091424413025378 + }, + { + "epoch": 1.541038525963149, + "flow_matching_loss": 0.030764632672071458, + "grad_norm": 0.2416084259748459, + "learning_rate": 2.855546199349214e-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": 0.01417373651638627, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.00662791975773871, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.03497539311647415, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.030764632672071458, + "weighted_rkd_loss": 0.017487696558237075 + }, + { + "epoch": 1.557788944723618, + "flow_matching_loss": 0.02624164503067732, + "grad_norm": 0.331834614276886, + "learning_rate": 2.85126876976474e-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": 0.013679224532097578, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.006975525338202715, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.03433397449553013, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.02624164503067732, + "weighted_rkd_loss": 0.017166987247765064 + }, + { + "epoch": 1.574539363484087, + "flow_matching_loss": 0.024381680972874166, + "grad_norm": 0.31837213039398193, + "learning_rate": 2.8469322313067726e-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": 0.013315047975629568, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.005786149669438601, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.03241624571382999, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.024381680972874166, + "weighted_rkd_loss": 0.016208122856914996 + }, + { + "epoch": 1.591289782244556, + "flow_matching_loss": 0.028025544807314874, + "grad_norm": 0.2301408052444458, + "learning_rate": 2.8425367736695522e-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": 0.012464443966746331, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.006268363306298852, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.031197250820696354, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.028025544807314874, + "weighted_rkd_loss": 0.015598625410348177 + }, + { + "epoch": 1.608040201005025, + "flow_matching_loss": 0.02433723080903292, + "grad_norm": 0.24389685690402985, + "learning_rate": 2.838082589124636e-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": 0.013607794232666493, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.006303474889136851, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.033519063331186774, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.02433723080903292, + "weighted_rkd_loss": 0.016759531665593387 + }, + { + "epoch": 1.624790619765494, + "flow_matching_loss": 0.022779523208737374, + "grad_norm": 0.28541818261146545, + "learning_rate": 2.8335698725124843e-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": 0.014848234411329032, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.006937592523172498, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.036634061485528946, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.022779523208737374, + "weighted_rkd_loss": 0.018317030742764473 + }, + { + "epoch": 1.641541038525963, + "flow_matching_loss": 0.023877214454114438, + "grad_norm": 0.2595161497592926, + "learning_rate": 2.8289988212339422e-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": 0.01390945939347148, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.006691856821998954, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.034510775655508044, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.023877214454114438, + "weighted_rkd_loss": 0.017255387827754022 + }, + { + "epoch": 1.658291457286432, + "flow_matching_loss": 0.022958588413894178, + "grad_norm": 0.27092212438583374, + "learning_rate": 2.8243696352415997e-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": 0.013638240005820989, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.006506198924034834, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.03378267884254456, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.022958588413894178, + "weighted_rkd_loss": 0.01689133942127228 + }, + { + "epoch": 1.675041876046901, + "flow_matching_loss": 0.027839783951640128, + "grad_norm": 0.2704690396785736, + "learning_rate": 2.8196825170310495e-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": 0.013532010465860366, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.006611468456685543, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.03367548901587725, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.027839783951640128, + "weighted_rkd_loss": 0.016837744507938623 + }, + { + "epoch": 1.69179229480737, + "flow_matching_loss": 0.02465656101703644, + "grad_norm": 0.2611840069293976, + "learning_rate": 2.8149376716320256e-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": 0.014256470091640949, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.007447991822846234, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.03596093207597732, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.02465656101703644, + "weighted_rkd_loss": 0.01798046603798866 + }, + { + "epoch": 1.708542713567839, + "flow_matching_loss": 0.02290959656238556, + "grad_norm": 0.23971238732337952, + "learning_rate": 2.8101353065994373e-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": 0.014609135687351227, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.007953010452911257, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.03717128206044436, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.02290959656238556, + "weighted_rkd_loss": 0.01858564103022218 + }, + { + "epoch": 1.725293132328308, + "flow_matching_loss": 0.02476640697568655, + "grad_norm": 0.28278639912605286, + "learning_rate": 2.8052756320042885e-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": 0.012388040497899055, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.005793318548239767, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.030569399893283843, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.02476640697568655, + "weighted_rkd_loss": 0.015284699946641922 + }, + { + "epoch": 1.742043551088777, + "flow_matching_loss": 0.022918140143156053, + "grad_norm": 0.24964776635169983, + "learning_rate": 2.8003588604244893e-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": 0.012319287657737732, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0058759675361216065, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.030514542944729328, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.022918140143156053, + "weighted_rkd_loss": 0.015257271472364664 + }, + { + "epoch": 1.758793969849246, + "flow_matching_loss": 0.024741691816598178, + "grad_norm": 0.2456982135772705, + "learning_rate": 2.7953852069355565e-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": 0.011992380395531655, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.007133734086528421, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.031118495017290117, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.024741691816598178, + "weighted_rkd_loss": 0.015559247508645058 + }, + { + "epoch": 1.775544388609715, + "flow_matching_loss": 0.02395295472815633, + "grad_norm": 0.2144646793603897, + "learning_rate": 2.790354889101206e-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": 0.01104045552201569, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.004955642647109926, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.02703655380755663, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.02395295472815633, + "weighted_rkd_loss": 0.013518276903778315 + }, + { + "epoch": 1.792294807370184, + "flow_matching_loss": 0.024379800911992787, + "grad_norm": 0.25276830792427063, + "learning_rate": 2.7852681269638366e-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": 0.012064874824136495, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.005810623452998698, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.02994037326425314, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.024379800911992787, + "weighted_rkd_loss": 0.01497018663212657 + }, + { + "epoch": 1.809045226130653, + "flow_matching_loss": 0.022948321886360645, + "grad_norm": 0.262026309967041, + "learning_rate": 2.780125143034902e-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": 0.011783893406391143, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.005772650917060673, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.029340437799692153, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.022948321886360645, + "weighted_rkd_loss": 0.014670218899846077 + }, + { + "epoch": 1.8257956448911221, + "flow_matching_loss": 0.025508361496031286, + "grad_norm": 0.2713301181793213, + "learning_rate": 2.7749261622851812e-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": 0.012106180377304553, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.005845029139891267, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.03005739003419876, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.025508361496031286, + "weighted_rkd_loss": 0.01502869501709938 + }, + { + "epoch": 1.8425460636515911, + "flow_matching_loss": 0.024567527882754802, + "grad_norm": 0.2601495385169983, + "learning_rate": 2.7696714121349335e-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": 0.011502832267433405, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.005032225488685071, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.028037889674305917, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.024567527882754802, + "weighted_rkd_loss": 0.014018944837152958 + }, + { + "epoch": 1.8592964824120601, + "flow_matching_loss": 0.02439136765897274, + "grad_norm": 0.2167050987482071, + "learning_rate": 2.764361122443954e-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": 0.011389804165810346, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.005811809515580535, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.028591417893767356, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.02439136765897274, + "weighted_rkd_loss": 0.014295708946883678 + }, + { + "epoch": 1.8760469011725294, + "flow_matching_loss": 0.021312791109085082, + "grad_norm": 0.2774104177951813, + "learning_rate": 2.758995525501517e-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": 0.012004026584327221, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.005796468770131469, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.029804521799087526, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.021312791109085082, + "weighted_rkd_loss": 0.014902260899543763 + }, + { + "epoch": 1.8927973199329984, + "flow_matching_loss": 0.02445399258285761, + "grad_norm": 0.2710288465023041, + "learning_rate": 2.753574856016215e-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": 0.01226965682581067, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.005771628674119711, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.03031094204634428, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.02445399258285761, + "weighted_rkd_loss": 0.01515547102317214 + }, + { + "epoch": 1.9095477386934674, + "flow_matching_loss": 0.022499824687838555, + "grad_norm": 0.22935397922992706, + "learning_rate": 2.7480993511056913e-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": 0.011755639128386975, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0051382229430601, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.02864950131624937, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.022499824687838555, + "weighted_rkd_loss": 0.014324750658124686 + }, + { + "epoch": 1.9262981574539364, + "flow_matching_loss": 0.021963876485824586, + "grad_norm": 0.21669921278953552, + "learning_rate": 2.7425692502862684e-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": 0.011520753987133503, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0047319270204752685, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.027773435041308404, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.021963876485824586, + "weighted_rkd_loss": 0.013886717520654202 + }, + { + "epoch": 1.9430485762144054, + "flow_matching_loss": 0.024818232655525206, + "grad_norm": 0.23156653344631195, + "learning_rate": 2.7369847954624728e-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": 0.011921488214284182, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0053805957548320295, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.029223571717739105, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.024818232655525206, + "weighted_rkd_loss": 0.014611785858869552 + }, + { + "epoch": 1.9597989949748744, + "flow_matching_loss": 0.0229154285043478, + "grad_norm": 0.2043662816286087, + "learning_rate": 2.731346230916448e-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": 0.011070125922560691, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.005242202361114323, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.0273824542760849, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.0229154285043478, + "weighted_rkd_loss": 0.01369122713804245 + }, + { + "epoch": 1.9765494137353434, + "flow_matching_loss": 0.022313889395445586, + "grad_norm": 0.2784120440483093, + "learning_rate": 2.7256538032972746e-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": 0.010846604779362678, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.004584192973561585, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.02627740241587162, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.022313889395445586, + "weighted_rkd_loss": 0.01313870120793581 + }, + { + "epoch": 1.9932998324958124, + "flow_matching_loss": 0.022565304301679134, + "grad_norm": 0.20291274785995483, + "learning_rate": 2.7199077616101773e-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": 0.010502919927239418, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0049030915135517715, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.02590893115848303, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.022565304301679134, + "weighted_rkd_loss": 0.012954465579241514 + }, + { + "epoch": 2.0100502512562812, + "flow_matching_loss": 0.02108859419822693, + "grad_norm": 0.2592752277851105, + "learning_rate": 2.714108357205634e-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": 0.011645914521068335, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.00536120398901403, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.02865303251892328, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.02108859419822693, + "weighted_rkd_loss": 0.01432651625946164 + }, + { + "epoch": 2.0268006700167502, + "flow_matching_loss": 0.02473500929772854, + "grad_norm": 0.2747367024421692, + "learning_rate": 2.7082558437683803e-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": 0.009690076718106866, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.004602333134971559, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.023982486315071584, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.02473500929772854, + "weighted_rkd_loss": 0.011991243157535792 + }, + { + "epoch": 2.0435510887772192, + "flow_matching_loss": 0.022570890467613937, + "grad_norm": 0.24187985062599182, + "learning_rate": 2.702350477306315e-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": 0.010515575297176838, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.004560630582273006, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.02559178089722991, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.022570890467613937, + "weighted_rkd_loss": 0.012795890448614955 + }, + { + "epoch": 2.0603015075376883, + "flow_matching_loss": 0.021877025440335273, + "grad_norm": 0.2488357424736023, + "learning_rate": 2.6963925161392963e-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": 0.009870412293821573, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.004250380373559892, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.023991205170750617, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.021877025440335273, + "weighted_rkd_loss": 0.011995602585375308 + }, + { + "epoch": 2.0770519262981573, + "flow_matching_loss": 0.022524043265730143, + "grad_norm": 0.24365104734897614, + "learning_rate": 2.6903822208878476e-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": 0.008939949795603752, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0040558589389547706, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.021935758553445338, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.022524043265730143, + "weighted_rkd_loss": 0.010967879276722669 + }, + { + "epoch": 2.0938023450586263, + "flow_matching_loss": 0.02082566823810339, + "grad_norm": 0.2463584691286087, + "learning_rate": 2.6843198544617535e-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": 0.012220189860090613, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.006351936119608581, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.030792316235601903, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.02082566823810339, + "weighted_rkd_loss": 0.015396158117800951 + }, + { + "epoch": 2.1105527638190953, + "flow_matching_loss": 0.02591631282120943, + "grad_norm": 0.22120581567287445, + "learning_rate": 2.6782056820485612e-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": 0.010837587667629123, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.004791317391209304, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.026466493122279645, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.02591631282120943, + "weighted_rkd_loss": 0.013233246561139823 + }, + { + "epoch": 2.1273031825795643, + "flow_matching_loss": 0.02475853394716978, + "grad_norm": 0.2626861035823822, + "learning_rate": 2.672039971101979e-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": 0.012477730866521597, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.00598516333848238, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.030940625071525573, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.02475853394716978, + "weighted_rkd_loss": 0.015470312535762787 + }, + { + "epoch": 2.1440536013400333, + "flow_matching_loss": 0.02421090826392174, + "grad_norm": 0.1990576982498169, + "learning_rate": 2.665822991330178e-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": 0.009951106877997518, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.004872644017450511, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.024774858076125384, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.02421090826392174, + "weighted_rkd_loss": 0.012387429038062692 + }, + { + "epoch": 2.1608040201005023, + "flow_matching_loss": 0.022640141285955907, + "grad_norm": 0.22442898154258728, + "learning_rate": 2.6595550146839934e-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": 0.011723242560401559, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.005248206853866577, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.02869469244033098, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.022640141285955907, + "weighted_rkd_loss": 0.01434734622016549 + }, + { + "epoch": 2.1775544388609713, + "flow_matching_loss": 0.023943004570901393, + "grad_norm": 0.3037925958633423, + "learning_rate": 2.6532363153450287e-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": 0.011830547265708447, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.005502653657458723, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.029163748025894165, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.023943004570901393, + "weighted_rkd_loss": 0.014581874012947083 + }, + { + "epoch": 2.1943048576214403, + "flow_matching_loss": 0.02036823332309723, + "grad_norm": 0.286495178937912, + "learning_rate": 2.6468671697136634e-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": 0.012905545998364687, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.006543920561671257, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.0323550121858716, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.02036823332309723, + "weighted_rkd_loss": 0.0161775060929358 + }, + { + "epoch": 2.2110552763819094, + "flow_matching_loss": 0.0228480763733387, + "grad_norm": 0.2888075113296509, + "learning_rate": 2.640447856396961e-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": 0.010347199253737926, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.005058633722364902, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.025753032229840756, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.0228480763733387, + "weighted_rkd_loss": 0.012876516114920378 + }, + { + "epoch": 2.2278056951423784, + "flow_matching_loss": 0.02162502771243453, + "grad_norm": 0.24059318006038666, + "learning_rate": 2.6339786561964805e-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": 0.010203090263530612, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.004981991858221591, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.025388172082602977, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.02162502771243453, + "weighted_rkd_loss": 0.012694086041301489 + }, + { + "epoch": 2.2445561139028474, + "flow_matching_loss": 0.019410558603703977, + "grad_norm": 0.2702176868915558, + "learning_rate": 2.6274598520959966e-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": 0.010505857458338142, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.004636408272199332, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.025648123398423196, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.019410558603703977, + "weighted_rkd_loss": 0.012824061699211598 + }, + { + "epoch": 2.2613065326633164, + "flow_matching_loss": 0.021643996052443982, + "grad_norm": 0.2329064905643463, + "learning_rate": 2.6208917292491185e-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": 0.010927976155653597, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.004669039882719517, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.026524991914629936, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.021643996052443982, + "weighted_rkd_loss": 0.013262495957314968 + }, + { + "epoch": 2.2780569514237854, + "flow_matching_loss": 0.02162062283605337, + "grad_norm": 0.23032520711421967, + "learning_rate": 2.6142745749668176e-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": 0.00848289243876934, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.004086268320679665, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.021052053198218346, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.02162062283605337, + "weighted_rkd_loss": 0.010526026599109173 + }, + { + "epoch": 2.2948073701842544, + "flow_matching_loss": 0.020955103635787963, + "grad_norm": 0.1977476179599762, + "learning_rate": 2.607608678704859e-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": 0.01016695871949196, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.004941023187711835, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.02527494076639414, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.020955103635787963, + "weighted_rkd_loss": 0.01263747038319707 + }, + { + "epoch": 2.3115577889447234, + "flow_matching_loss": 0.02056379960849881, + "grad_norm": 0.2108568698167801, + "learning_rate": 2.600894332051139e-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": 0.01030728747136891, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.004301543720066548, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.024916118942201138, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.02056379960849881, + "weighted_rkd_loss": 0.012458059471100569 + }, + { + "epoch": 2.3283082077051924, + "flow_matching_loss": 0.025450243800878524, + "grad_norm": 0.23199211061000824, + "learning_rate": 2.594131828712932e-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": 0.00888884300366044, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.00437500849366188, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.02215269412845373, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.025450243800878524, + "weighted_rkd_loss": 0.011076347064226865 + }, + { + "epoch": 2.3450586264656614, + "flow_matching_loss": 0.021527566574513914, + "grad_norm": 0.29200154542922974, + "learning_rate": 2.5873214645040414e-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": 0.010741008073091507, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.005127520440146327, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.026609536819159986, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.021527566574513914, + "weighted_rkd_loss": 0.013304768409579993 + }, + { + "epoch": 2.3618090452261304, + "flow_matching_loss": 0.021938143484294414, + "grad_norm": 0.2158461958169937, + "learning_rate": 2.5804635373318604e-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": 0.011550693772733212, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.005129396705888212, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.028230783902108668, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.021938143484294414, + "weighted_rkd_loss": 0.014115391951054334 + }, + { + "epoch": 2.3785594639865995, + "flow_matching_loss": 0.022317277640104292, + "grad_norm": 0.2570502460002899, + "learning_rate": 2.57355834718434e-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": 0.008357560494914651, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.003908583358861506, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.02062370451167226, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.022317277640104292, + "weighted_rkd_loss": 0.01031185225583613 + }, + { + "epoch": 2.3953098827470685, + "flow_matching_loss": 0.02067250395193696, + "grad_norm": 0.23166990280151367, + "learning_rate": 2.566606196116867e-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": 0.010661517130210995, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.005037658615037799, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.02636069282889366, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.02067250395193696, + "weighted_rkd_loss": 0.01318034641444683 + }, + { + "epoch": 2.4120603015075375, + "flow_matching_loss": 0.022506509348750113, + "grad_norm": 0.23834720253944397, + "learning_rate": 2.559607388239051e-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": 0.00992490230128169, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.004856222122907638, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.024706026911735533, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.022506509348750113, + "weighted_rkd_loss": 0.012353013455867767 + }, + { + "epoch": 2.4288107202680065, + "flow_matching_loss": 0.019241156708449125, + "grad_norm": 0.2729126811027527, + "learning_rate": 2.552562229701422e-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": 0.010186821967363358, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.004899685806594789, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.025273329950869082, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.019241156708449125, + "weighted_rkd_loss": 0.012636664975434541 + }, + { + "epoch": 2.4455611390284755, + "flow_matching_loss": 0.023337592743337153, + "grad_norm": 0.19402694702148438, + "learning_rate": 2.545471028682036e-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": 0.009790295595303177, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.004415846662595868, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.023996437899768353, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.023337592743337153, + "weighted_rkd_loss": 0.011998218949884176 + }, + { + "epoch": 2.4623115577889445, + "flow_matching_loss": 0.02342185452580452, + "grad_norm": 0.26485997438430786, + "learning_rate": 2.538334095373e-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": 0.009926773328334092, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.004456087900325656, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.02430963460355997, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.02342185452580452, + "weighted_rkd_loss": 0.012154817301779985 + }, + { + "epoch": 2.4790619765494135, + "flow_matching_loss": 0.022115869354456662, + "grad_norm": 0.21239666640758514, + "learning_rate": 2.531151741966896e-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": 0.008159320428967476, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0040833617094904184, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.020402002427726985, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.022115869354456662, + "weighted_rkd_loss": 0.010201001213863492 + }, + { + "epoch": 2.4958123953098825, + "flow_matching_loss": 0.022036940045654773, + "grad_norm": 0.2418198585510254, + "learning_rate": 2.5239242826431302e-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": 0.008541764924302697, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.003945641336031258, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.021029170975089072, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.022036940045654773, + "weighted_rkd_loss": 0.010514585487544536 + }, + { + "epoch": 2.5125628140703515, + "flow_matching_loss": 0.021928760688751936, + "grad_norm": 0.21864691376686096, + "learning_rate": 2.5166520335541865e-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": 0.009622622048482299, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.004568752995692193, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.02381399692967534, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.021928760688751936, + "weighted_rkd_loss": 0.01190699846483767 + }, + { + "epoch": 2.5293132328308205, + "flow_matching_loss": 0.02287863790988922, + "grad_norm": 0.24944423139095306, + "learning_rate": 2.5093353128117996e-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": 0.00825018659234047, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0038231824059039355, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.02032355535775423, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.02287863790988922, + "weighted_rkd_loss": 0.010161777678877115 + }, + { + "epoch": 2.5460636515912896, + "flow_matching_loss": 0.02119820462539792, + "grad_norm": 0.19911466538906097, + "learning_rate": 2.501974440473037e-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": 0.010588160343468189, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0056390015874058005, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.02681532232090831, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.02119820462539792, + "weighted_rkd_loss": 0.013407661160454155 + }, + { + "epoch": 2.5628140703517586, + "flow_matching_loss": 0.022425088565796614, + "grad_norm": 0.21942198276519775, + "learning_rate": 2.4945697385263015e-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": 0.008955000853165983, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.004014491592533887, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.021924493461847307, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.022425088565796614, + "weighted_rkd_loss": 0.010962246730923653 + }, + { + "epoch": 2.5795644891122276, + "flow_matching_loss": 0.020732478983700276, + "grad_norm": 0.18362075090408325, + "learning_rate": 2.487121530877243e-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": 0.00954368463717401, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.004505707276985049, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.023593076318502427, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.020732478983700276, + "weighted_rkd_loss": 0.011796538159251213 + }, + { + "epoch": 2.5963149078726966, + "flow_matching_loss": 0.019448608718812467, + "grad_norm": 0.22332197427749634, + "learning_rate": 2.4796301433345943e-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": 0.009949681954458356, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.004817895125597715, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.024717259034514428, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.019448608718812467, + "weighted_rkd_loss": 0.012358629517257214 + }, + { + "epoch": 2.6130653266331656, + "flow_matching_loss": 0.020631531719118357, + "grad_norm": 0.20138533413410187, + "learning_rate": 2.4720959035959148e-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": 0.00828108130954206, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.004354966944083572, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.020917129516601563, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.020631531719118357, + "weighted_rkd_loss": 0.010458564758300782 + }, + { + "epoch": 2.6298157453936346, + "flow_matching_loss": 0.019199035596102476, + "grad_norm": 0.22796088457107544, + "learning_rate": 2.464519141233257e-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": 0.008544412907212973, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.003954147803597152, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.021042973920702934, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.019199035596102476, + "weighted_rkd_loss": 0.010521486960351467 + }, + { + "epoch": 2.6465661641541036, + "flow_matching_loss": 0.020341791026294232, + "grad_norm": 0.22180351614952087, + "learning_rate": 2.456900187678754e-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": 0.00856758183799684, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0038478421280160545, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.020983005687594414, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.020341791026294232, + "weighted_rkd_loss": 0.010491502843797207 + }, + { + "epoch": 2.6633165829145726, + "flow_matching_loss": 0.018619582895189522, + "grad_norm": 0.21050786972045898, + "learning_rate": 2.4492393762101147e-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": 0.009519612230360509, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.004517683316953481, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.023556907661259174, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.018619582895189522, + "weighted_rkd_loss": 0.011778453830629587 + }, + { + "epoch": 2.6800670016750416, + "flow_matching_loss": 0.01987162446603179, + "grad_norm": 0.21779532730579376, + "learning_rate": 2.4415370419360508e-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": 0.009196092328056692, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0039106435142457485, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.022302828263491392, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.01987162446603179, + "weighted_rkd_loss": 0.011151414131745696 + }, + { + "epoch": 2.6968174204355106, + "flow_matching_loss": 0.01982830436900258, + "grad_norm": 0.20559139549732208, + "learning_rate": 2.4337935217816157e-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": 0.010465317685157061, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.004926157207228244, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.025856792367994784, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.01982830436900258, + "weighted_rkd_loss": 0.012928396183997392 + }, + { + "epoch": 2.7135678391959797, + "flow_matching_loss": 0.024718323722481728, + "grad_norm": 0.20431019365787506, + "learning_rate": 2.4260091544734665e-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": 0.010142649663612247, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.00459180234465748, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.024877101369202136, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.024718323722481728, + "weighted_rkd_loss": 0.012438550684601068 + }, + { + "epoch": 2.7303182579564487, + "flow_matching_loss": 0.022330477088689803, + "grad_norm": 0.21411430835723877, + "learning_rate": 2.4181842805250466e-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": 0.008193683298304676, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0040412382688373326, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.0204286047257483, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.022330477088689803, + "weighted_rkd_loss": 0.01021430236287415 + }, + { + "epoch": 2.7470686767169177, + "flow_matching_loss": 0.01971994461491704, + "grad_norm": 0.19193406403064728, + "learning_rate": 2.4103192422216923e-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": 0.009616542607545853, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.004992820438928902, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.024225905537605286, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.01971994461491704, + "weighted_rkd_loss": 0.012112952768802643 + }, + { + "epoch": 2.7638190954773867, + "flow_matching_loss": 0.020513110049068927, + "grad_norm": 0.2666885256767273, + "learning_rate": 2.402414383605658e-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": 0.009339694399386644, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.004481942718848586, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.023161331377923488, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.020513110049068927, + "weighted_rkd_loss": 0.011580665688961744 + }, + { + "epoch": 2.7805695142378557, + "flow_matching_loss": 0.019615318812429904, + "grad_norm": 0.23744481801986694, + "learning_rate": 2.3944700504610674e-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": 0.010119090462103486, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.004129025409929455, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.024367206171154975, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.019615318812429904, + "weighted_rkd_loss": 0.012183603085577487 + }, + { + "epoch": 2.7973199329983247, + "flow_matching_loss": 0.019463876076042653, + "grad_norm": 0.24974969029426575, + "learning_rate": 2.3864865902987883e-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": 0.008990717539563774, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0040746829472482204, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.022056118212640287, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.019463876076042653, + "weighted_rkd_loss": 0.011028059106320143 + }, + { + "epoch": 2.8140703517587937, + "flow_matching_loss": 0.02275766897946596, + "grad_norm": 0.2230592966079712, + "learning_rate": 2.378464352341231e-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": 0.009455362567678094, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.004774155165068805, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.023684880696237087, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.02275766897946596, + "weighted_rkd_loss": 0.011842440348118544 + }, + { + "epoch": 2.8308207705192627, + "flow_matching_loss": 0.01860237428918481, + "grad_norm": 0.22969353199005127, + "learning_rate": 2.3704036875070728e-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": 0.008334634220227599, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.003853255487047136, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.020522524043917655, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.01860237428918481, + "weighted_rkd_loss": 0.010261262021958827 + }, + { + "epoch": 2.8475711892797317, + "flow_matching_loss": 0.022035404108464718, + "grad_norm": 0.19043883681297302, + "learning_rate": 2.362304948395906e-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": 0.009156103385612369, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.004319961508736014, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.022632168233394624, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.022035404108464718, + "weighted_rkd_loss": 0.011316084116697312 + }, + { + "epoch": 2.8643216080402008, + "flow_matching_loss": 0.020399292930960655, + "grad_norm": 0.1880931854248047, + "learning_rate": 2.3541684892728157e-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": 0.008248971542343497, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.004034064570441842, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.02053200760856271, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.020399292930960655, + "weighted_rkd_loss": 0.010266003804281354 + }, + { + "epoch": 2.8810720268006698, + "flow_matching_loss": 0.023524169996380805, + "grad_norm": 0.21712803840637207, + "learning_rate": 2.345994666052883e-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": 0.008792503830045461, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.004066857602447271, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.021651864983141422, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.023524169996380805, + "weighted_rkd_loss": 0.010825932491570711 + }, + { + "epoch": 2.8978224455611388, + "flow_matching_loss": 0.018634811509400606, + "grad_norm": 0.17355115711688995, + "learning_rate": 2.337783836285614e-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": 0.0091339401435107, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.003754345467314124, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.022022225055843593, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.018634811509400606, + "weighted_rkd_loss": 0.011011112527921797 + }, + { + "epoch": 2.914572864321608, + "flow_matching_loss": 0.01859223246574402, + "grad_norm": 0.2472020834684372, + "learning_rate": 2.3295363591393026e-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": 0.00817334484308958, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0036398014053702354, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01998649090528488, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.01859223246574402, + "weighted_rkd_loss": 0.00999324545264244 + }, + { + "epoch": 2.931323283082077, + "flow_matching_loss": 0.020342006627470256, + "grad_norm": 0.2341446578502655, + "learning_rate": 2.3212525953853174e-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": 0.00935922097414732, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0042345151770859955, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.022952956706285478, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.020342006627470256, + "weighted_rkd_loss": 0.011476478353142739 + }, + { + "epoch": 2.948073701842546, + "flow_matching_loss": 0.01856496771797538, + "grad_norm": 0.23332548141479492, + "learning_rate": 2.3129329073823206e-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": 0.00806609825231135, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0038775680121034386, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.02000976465642452, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.01856496771797538, + "weighted_rkd_loss": 0.01000488232821226 + }, + { + "epoch": 2.964824120603015, + "flow_matching_loss": 0.020582755748182534, + "grad_norm": 0.22185544669628143, + "learning_rate": 2.3045776590604175e-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": 0.010403018863871693, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.004640304553322494, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.025446342397481204, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.020582755748182534, + "weighted_rkd_loss": 0.012723171198740602 + }, + { + "epoch": 2.981574539363484, + "flow_matching_loss": 0.020155472122132777, + "grad_norm": 0.2461974173784256, + "learning_rate": 2.296187215905237e-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": 0.00791851249523461, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0035846429644152524, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.019421668257564306, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.020155472122132777, + "weighted_rkd_loss": 0.009710834128782153 + }, + { + "epoch": 2.998324958123953, + "flow_matching_loss": 0.0227069896645844, + "grad_norm": 0.2363586276769638, + "learning_rate": 2.2877619449419437e-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": 0.008106299256905914, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.00372429087292403, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.019936889316886662, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.0227069896645844, + "weighted_rkd_loss": 0.009968444658443331 + }, + { + "epoch": 3.0150753768844223, + "flow_matching_loss": 0.01800302853807807, + "grad_norm": 0.2395818680524826, + "learning_rate": 2.2793022147191843e-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": 0.008237518183887006, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.004040196281857788, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.020515232533216476, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.01800302853807807, + "weighted_rkd_loss": 0.010257616266608238 + }, + { + "epoch": 3.0318257956448913, + "flow_matching_loss": 0.019564515352249144, + "grad_norm": 0.22039702534675598, + "learning_rate": 2.2708083952929634e-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": 0.008851285045966505, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.003621540090534836, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.021324109751731157, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.019564515352249144, + "weighted_rkd_loss": 0.010662054875865579 + }, + { + "epoch": 3.0485762144053603, + "flow_matching_loss": 0.02003264194354415, + "grad_norm": 0.22414322197437286, + "learning_rate": 2.2622808582104594e-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": 0.006905040238052607, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.003053883952088654, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01686396459117532, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.02003264194354415, + "weighted_rkd_loss": 0.00843198229558766 + }, + { + "epoch": 3.0653266331658293, + "flow_matching_loss": 0.018170845229178667, + "grad_norm": 0.22030091285705566, + "learning_rate": 2.2537199764937703e-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": 0.007743219006806612, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.003740711836144328, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.019227149710059167, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.018170845229178667, + "weighted_rkd_loss": 0.009613574855029583 + }, + { + "epoch": 3.0820770519262983, + "flow_matching_loss": 0.019879622850567102, + "grad_norm": 0.1730256974697113, + "learning_rate": 2.245126124623595e-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": 0.008711709873750806, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0041843382874503735, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.021607757732272147, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.019879622850567102, + "weighted_rkd_loss": 0.010803878866136074 + }, + { + "epoch": 3.0988274706867673, + "flow_matching_loss": 0.01942541915923357, + "grad_norm": 0.15700264275074005, + "learning_rate": 2.2364996785228547e-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": 0.00775566566735506, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0036699424032121896, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.019181273877620697, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.01942541915923357, + "weighted_rkd_loss": 0.009590636938810349 + }, + { + "epoch": 3.1155778894472363, + "flow_matching_loss": 0.019799060095101594, + "grad_norm": 0.19989314675331116, + "learning_rate": 2.227841015540248e-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": 0.008485706709325313, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.003620952763594687, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.020592366252094506, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.019799060095101594, + "weighted_rkd_loss": 0.010296183126047253 + }, + { + "epoch": 3.1323283082077054, + "flow_matching_loss": 0.020041495002806185, + "grad_norm": 0.24366852641105652, + "learning_rate": 2.2191505144337433e-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": 0.008763994136825205, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0038334209937602283, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.021361409313976766, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.020041495002806185, + "weighted_rkd_loss": 0.010680704656988383 + }, + { + "epoch": 3.1490787269681744, + "flow_matching_loss": 0.019042767770588397, + "grad_norm": 0.22834265232086182, + "learning_rate": 2.210428555354013e-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": 0.007715944433584809, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.003302133153192699, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01873402213677764, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.019042767770588397, + "weighted_rkd_loss": 0.00936701106838882 + }, + { + "epoch": 3.1658291457286434, + "flow_matching_loss": 0.018769701849669217, + "grad_norm": 0.2096819430589676, + "learning_rate": 2.201675519827802e-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": 0.008365185000002384, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0035678216489031913, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.020298191532492637, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.018769701849669217, + "weighted_rkd_loss": 0.010149095766246318 + }, + { + "epoch": 3.1825795644891124, + "flow_matching_loss": 0.018232905119657517, + "grad_norm": 0.17371851205825806, + "learning_rate": 2.1928917907412403e-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": 0.008268481539562344, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.003637303062714636, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.020174266118556262, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.018232905119657517, + "weighted_rkd_loss": 0.010087133059278131 + }, + { + "epoch": 3.1993299832495814, + "flow_matching_loss": 0.019946189783513545, + "grad_norm": 0.27510249614715576, + "learning_rate": 2.1840777523230937e-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": 0.008179319882765412, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.003651675768196583, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.02001031571999192, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.019946189783513545, + "weighted_rkd_loss": 0.01000515785999596 + }, + { + "epoch": 3.2160804020100504, + "flow_matching_loss": 0.019676271919161083, + "grad_norm": 0.19415171444416046, + "learning_rate": 2.1752337901279552e-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": 0.007942137494683265, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.003817263268865645, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.019701538234949113, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.019676271919161083, + "weighted_rkd_loss": 0.009850769117474557 + }, + { + "epoch": 3.2328308207705194, + "flow_matching_loss": 0.01967736016958952, + "grad_norm": 0.21885381639003754, + "learning_rate": 2.166360291019383e-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": 0.007890013046562672, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.003441098565235734, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.019221124611794947, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.01967736016958952, + "weighted_rkd_loss": 0.009610562305897474 + }, + { + "epoch": 3.2495812395309884, + "flow_matching_loss": 0.01775270150974393, + "grad_norm": 0.19905905425548553, + "learning_rate": 2.157457643152973e-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": 0.0073103581089526415, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.003515601658727974, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.0181363177485764, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.01775270150974393, + "weighted_rkd_loss": 0.0090681588742882 + }, + { + "epoch": 3.2663316582914574, + "flow_matching_loss": 0.02068208493292332, + "grad_norm": 0.20610065758228302, + "learning_rate": 2.148526235959385e-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": 0.008260679943487049, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0035158348036929965, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.020037194434553384, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.02068208493292332, + "weighted_rkd_loss": 0.010018597217276692 + }, + { + "epoch": 3.2830820770519265, + "flow_matching_loss": 0.019399074651300908, + "grad_norm": 0.21385829150676727, + "learning_rate": 2.1395664601273016e-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": 0.00821441924199462, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0036132696783170103, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.02004210827872157, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.019399074651300908, + "weighted_rkd_loss": 0.010021054139360785 + }, + { + "epoch": 3.2998324958123955, + "flow_matching_loss": 0.019065764639526604, + "grad_norm": 0.19990454614162445, + "learning_rate": 2.130578707586345e-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": 0.008452633442357183, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.004286817531101406, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.021192084439098834, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.019065764639526604, + "weighted_rkd_loss": 0.010596042219549417 + }, + { + "epoch": 3.3165829145728645, + "flow_matching_loss": 0.020254147704690695, + "grad_norm": 0.18197959661483765, + "learning_rate": 2.1215633714899264e-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": 0.007969699334353209, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0034585506888106464, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01939794924110174, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.020254147704690695, + "weighted_rkd_loss": 0.00969897462055087 + }, + { + "epoch": 3.3333333333333335, + "flow_matching_loss": 0.019262891449034213, + "grad_norm": 0.1941157728433609, + "learning_rate": 2.112520846198053e-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": 0.007744752708822489, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.003361690975725651, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.018851196579635143, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.019262891449034213, + "weighted_rkd_loss": 0.009425598289817571 + }, + { + "epoch": 3.3500837520938025, + "flow_matching_loss": 0.020645213034003972, + "grad_norm": 0.24348419904708862, + "learning_rate": 2.103451527260074e-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": 0.008306558383628726, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0033074723556637763, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.019920589029788972, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.020645213034003972, + "weighted_rkd_loss": 0.009960294514894486 + }, + { + "epoch": 3.3668341708542715, + "flow_matching_loss": 0.01771964505314827, + "grad_norm": 0.2650719881057739, + "learning_rate": 2.094355811397381e-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": 0.00872551086358726, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0037666106363758445, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.021217632200568916, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.01771964505314827, + "weighted_rkd_loss": 0.010608816100284458 + }, + { + "epoch": 3.3835845896147405, + "flow_matching_loss": 0.02218597140163183, + "grad_norm": 0.2130243331193924, + "learning_rate": 2.0852340964860522e-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": 0.008654556702822447, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0042005166411399845, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.021509630233049394, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.02218597140163183, + "weighted_rkd_loss": 0.010754815116524697 + }, + { + "epoch": 3.4003350083752095, + "flow_matching_loss": 0.01848809625953436, + "grad_norm": 0.17124658823013306, + "learning_rate": 2.076086781539447e-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": 0.007023377157747745, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.002872368926182389, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.016919123008847237, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.01848809625953436, + "weighted_rkd_loss": 0.008459561504423618 + }, + { + "epoch": 3.4170854271356785, + "flow_matching_loss": 0.017233775090426208, + "grad_norm": 0.2589413523674011, + "learning_rate": 2.066914266690755e-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": 0.008169579645618796, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0034815609222278, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.019820720050483943, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.017233775090426208, + "weighted_rkd_loss": 0.009910360025241972 + }, + { + "epoch": 3.4338358458961475, + "flow_matching_loss": 0.017476903088390828, + "grad_norm": 0.195095032453537, + "learning_rate": 2.0577169531754906e-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": 0.008019447745755314, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.003486019931733608, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.019524915516376494, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.017476903088390828, + "weighted_rkd_loss": 0.009762457758188247 + }, + { + "epoch": 3.4505862646566166, + "flow_matching_loss": 0.020145728811621667, + "grad_norm": 0.15870703756809235, + "learning_rate": 2.048495243313942e-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": 0.006887005269527435, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0027772085042670367, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.016551219020038845, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.020145728811621667, + "weighted_rkd_loss": 0.008275609510019422 + }, + { + "epoch": 3.4673366834170856, + "flow_matching_loss": 0.017395683471113442, + "grad_norm": 0.18425144255161285, + "learning_rate": 2.039249540493572e-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": 0.0067982316017150875, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.002812152076512575, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.016408615000545978, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.017395683471113442, + "weighted_rkd_loss": 0.008204307500272989 + }, + { + "epoch": 3.4840871021775546, + "flow_matching_loss": 0.017085911240428686, + "grad_norm": 0.24657684564590454, + "learning_rate": 2.0299802491513755e-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": 0.00758605794981122, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0035657767555676402, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.018737892620265484, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.017085911240428686, + "weighted_rkd_loss": 0.009368946310132742 + }, + { + "epoch": 3.5008375209380236, + "flow_matching_loss": 0.01852244548499584, + "grad_norm": 0.2186572253704071, + "learning_rate": 2.0206877747561828e-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": 0.008175423974171281, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.004138233431149274, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.020489081274718046, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.01852244548499584, + "weighted_rkd_loss": 0.010244540637359023 + }, + { + "epoch": 3.5175879396984926, + "flow_matching_loss": 0.017001311294734478, + "grad_norm": 0.22371233999729156, + "learning_rate": 2.011372523790927e-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": 0.008435172354802489, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.003978886944241822, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.020849231723695995, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.017001311294734478, + "weighted_rkd_loss": 0.010424615861847997 + }, + { + "epoch": 3.5343383584589616, + "flow_matching_loss": 0.016847202088683844, + "grad_norm": 0.19801795482635498, + "learning_rate": 2.0020349037348643e-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": 0.008030201774090528, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0033957967069000007, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.019456200394779445, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.016847202088683844, + "weighted_rkd_loss": 0.009728100197389723 + }, + { + "epoch": 3.5510887772194306, + "flow_matching_loss": 0.01707222471013665, + "grad_norm": 0.20372146368026733, + "learning_rate": 1.9926753230457443e-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": 0.008735480671748518, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0035827487939968706, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.021053710021078587, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.01707222471013665, + "weighted_rkd_loss": 0.010526855010539293 + }, + { + "epoch": 3.5678391959798996, + "flow_matching_loss": 0.02113640746101737, + "grad_norm": 0.2022916078567505, + "learning_rate": 1.983294191141948e-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": 0.006674180738627911, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0029194900183938445, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01626785146072507, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.02113640746101737, + "weighted_rkd_loss": 0.008133925730362535 + }, + { + "epoch": 3.5845896147403686, + "flow_matching_loss": 0.019594037998467685, + "grad_norm": 0.19287943840026855, + "learning_rate": 1.973891918384575e-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": 0.008663790998980403, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.004071521968580782, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.021399103943258523, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.019594037998467685, + "weighted_rkd_loss": 0.010699551971629261 + }, + { + "epoch": 3.6013400335008376, + "flow_matching_loss": 0.02062170347198844, + "grad_norm": 0.21747611463069916, + "learning_rate": 1.9644689160594945e-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": 0.00737433722242713, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0035712956450879574, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.018319970276206732, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.02062170347198844, + "weighted_rkd_loss": 0.009159985138103366 + }, + { + "epoch": 3.6180904522613067, + "flow_matching_loss": 0.01748494440689683, + "grad_norm": 0.23034244775772095, + "learning_rate": 1.9550255963593553e-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": 0.007272516191005707, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0033656290732324125, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.017910661175847054, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.01748494440689683, + "weighted_rkd_loss": 0.008955330587923527 + }, + { + "epoch": 3.6348408710217757, + "flow_matching_loss": 0.018562578503042458, + "grad_norm": 0.18812276422977448, + "learning_rate": 1.9455623723655524e-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": 0.006702336249873042, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.003232569736428559, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01663724211975932, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.018562578503042458, + "weighted_rkd_loss": 0.00831862105987966 + }, + { + "epoch": 3.6515912897822447, + "flow_matching_loss": 0.019608804117888214, + "grad_norm": 0.16253460943698883, + "learning_rate": 1.9360796580301606e-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": 0.0070120657328516245, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0028424362768419086, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01686656791716814, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.019608804117888214, + "weighted_rkd_loss": 0.00843328395858407 + }, + { + "epoch": 3.6683417085427137, + "flow_matching_loss": 0.019909032620489597, + "grad_norm": 0.1944153606891632, + "learning_rate": 1.9265778681578245e-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": 0.007547950092703104, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.003626455180346966, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.018722355365753174, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.019909032620489597, + "weighted_rkd_loss": 0.009361177682876587 + }, + { + "epoch": 3.6850921273031827, + "flow_matching_loss": 0.017008159402757884, + "grad_norm": 0.24014660716056824, + "learning_rate": 1.9170574183876143e-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": 0.00804790472611785, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0036449329694733025, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.019740742444992066, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.017008159402757884, + "weighted_rkd_loss": 0.009870371222496033 + }, + { + "epoch": 3.7018425460636517, + "flow_matching_loss": 0.015542886033654214, + "grad_norm": 0.19243793189525604, + "learning_rate": 1.9075187251748452e-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": 0.00820448799058795, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0038957787677645684, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.020304754748940466, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.015542886033654214, + "weighted_rkd_loss": 0.010152377374470233 + }, + { + "epoch": 3.7185929648241207, + "flow_matching_loss": 0.018036157917231323, + "grad_norm": 0.19585657119750977, + "learning_rate": 1.8979622057728596e-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": 0.006451038084924221, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0027595918742008506, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.015661667939275502, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.018036157917231323, + "weighted_rkd_loss": 0.007830833969637751 + }, + { + "epoch": 3.7353433835845897, + "flow_matching_loss": 0.018009207025170327, + "grad_norm": 0.24362027645111084, + "learning_rate": 1.8883882782147748e-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": 0.007770774234086275, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0031818384886719286, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01872338680550456, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.018009207025170327, + "weighted_rkd_loss": 0.00936169340275228 + }, + { + "epoch": 3.7520938023450587, + "flow_matching_loss": 0.019524939823895693, + "grad_norm": 0.2122136652469635, + "learning_rate": 1.8787973612951967e-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": 0.007774351211264729, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0034809625474736096, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.019029664900153877, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.019524939823895693, + "weighted_rkd_loss": 0.009514832450076938 + }, + { + "epoch": 3.7688442211055277, + "flow_matching_loss": 0.018598716892302036, + "grad_norm": 0.1975695788860321, + "learning_rate": 1.8691898745519028e-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": 0.006873442605137825, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.002597982867155224, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01634486811235547, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.018598716892302036, + "weighted_rkd_loss": 0.008172434056177735 + }, + { + "epoch": 3.7855946398659968, + "flow_matching_loss": 0.020560270361602306, + "grad_norm": 0.1811131238937378, + "learning_rate": 1.8595662382474863e-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": 0.00680540797766298, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.002947376715019345, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.016558192390948533, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.020560270361602306, + "weighted_rkd_loss": 0.008279096195474267 + }, + { + "epoch": 3.8023450586264658, + "flow_matching_loss": 0.01634566029533744, + "grad_norm": 0.1560610830783844, + "learning_rate": 1.8499268733509766e-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": 0.007181188557296991, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0033734793891198934, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.017735856398940087, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.01634566029533744, + "weighted_rkd_loss": 0.008867928199470044 + }, + { + "epoch": 3.819095477386935, + "flow_matching_loss": 0.01920417882502079, + "grad_norm": 0.1946534514427185, + "learning_rate": 1.84027220151942e-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": 0.006485169334337115, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0032176646520383654, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.016188003495335578, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.01920417882502079, + "weighted_rkd_loss": 0.008094001747667789 + }, + { + "epoch": 3.835845896147404, + "flow_matching_loss": 0.018274603225290776, + "grad_norm": 0.2120642513036728, + "learning_rate": 1.83060264507944e-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": 0.007450972450897098, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0033969814423471687, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.018298926390707494, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.018274603225290776, + "weighted_rkd_loss": 0.009149463195353747 + }, + { + "epoch": 3.852596314907873, + "flow_matching_loss": 0.017578314803540707, + "grad_norm": 0.19296100735664368, + "learning_rate": 1.82091862700876e-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": 0.005799337150529027, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0024415326886810364, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.014040206838399173, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.017578314803540707, + "weighted_rkd_loss": 0.007020103419199586 + }, + { + "epoch": 3.869346733668342, + "flow_matching_loss": 0.018563490454107524, + "grad_norm": 0.205954447388649, + "learning_rate": 1.8112205709177023e-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": 0.006602625688537955, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0030158299836330117, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.016221081465482713, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.018563490454107524, + "weighted_rkd_loss": 0.008110540732741357 + }, + { + "epoch": 3.886097152428811, + "flow_matching_loss": 0.01872290503233671, + "grad_norm": 0.20086728036403656, + "learning_rate": 1.8015089010306585e-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": 0.006772564933635294, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0029774153837934135, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01652254508808255, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.01872290503233671, + "weighted_rkd_loss": 0.008261272544041276 + }, + { + "epoch": 3.90284757118928, + "flow_matching_loss": 0.018443399760872125, + "grad_norm": 0.2528753876686096, + "learning_rate": 1.7917840421675298e-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": 0.006297580199316144, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.002838477923069149, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.015433638356626034, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.018443399760872125, + "weighted_rkd_loss": 0.007716819178313017 + }, + { + "epoch": 3.919597989949749, + "flow_matching_loss": 0.016544591914862395, + "grad_norm": 0.1990911215543747, + "learning_rate": 1.7820464197251474e-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": 0.005723108258098364, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.002620975370518863, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01406719172373414, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.016544591914862395, + "weighted_rkd_loss": 0.00703359586186707 + }, + { + "epoch": 3.936348408710218, + "flow_matching_loss": 0.02259768210351467, + "grad_norm": 0.21213243901729584, + "learning_rate": 1.7722964596586627e-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": 0.006221555732190609, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.002580338914413005, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.015023450460284948, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.02259768210351467, + "weighted_rkd_loss": 0.007511725230142474 + }, + { + "epoch": 3.953098827470687, + "flow_matching_loss": 0.018023871071636676, + "grad_norm": 0.21245522797107697, + "learning_rate": 1.7625345884629144e-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": 0.007134971069172024, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0029236291884444654, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01719357119873166, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.018023871071636676, + "weighted_rkd_loss": 0.00859678559936583 + }, + { + "epoch": 3.969849246231156, + "flow_matching_loss": 0.017593097407370806, + "grad_norm": 0.17448540031909943, + "learning_rate": 1.7527612331537722e-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": 0.0060433610808104275, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.002813521446660161, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.014900243375450373, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.017593097407370806, + "weighted_rkd_loss": 0.007450121687725186 + }, + { + "epoch": 3.986599664991625, + "flow_matching_loss": 0.01726466277614236, + "grad_norm": 0.2096206098794937, + "learning_rate": 1.7429768212494593e-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": 0.006471162941306829, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0029486360028386114, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.015890961978584528, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.01726466277614236, + "weighted_rkd_loss": 0.007945480989292264 + }, + { + "epoch": 4.0033500837520934, + "flow_matching_loss": 0.017108506895601748, + "grad_norm": 0.24649129807949066, + "learning_rate": 1.7331817807518497e-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": 0.008349902275949717, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0036193238804116843, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.020319128409028053, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.017108506895601748, + "weighted_rkd_loss": 0.010159564204514027 + }, + { + "epoch": 4.0201005025125625, + "flow_matching_loss": 0.017888367548584937, + "grad_norm": 0.16043002903461456, + "learning_rate": 1.723376540127745e-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": 0.006572570186108351, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0026901500299572946, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01583529030904174, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.017888367548584937, + "weighted_rkd_loss": 0.00791764515452087 + }, + { + "epoch": 4.0368509212730315, + "flow_matching_loss": 0.01659221211448312, + "grad_norm": 0.2388010323047638, + "learning_rate": 1.7135615282901357e-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": 0.0061241401825100185, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0029370995704084634, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.015185380168259145, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.01659221211448312, + "weighted_rkd_loss": 0.007592690084129572 + }, + { + "epoch": 4.0536013400335005, + "flow_matching_loss": 0.01957662058994174, + "grad_norm": 0.18511873483657837, + "learning_rate": 1.7037371745794358e-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": 0.005709731136448681, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.00265506012365222, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.014074522443115712, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.01957662058994174, + "weighted_rkd_loss": 0.007037261221557856 + }, + { + "epoch": 4.0703517587939695, + "flow_matching_loss": 0.01813545823097229, + "grad_norm": 0.19960571825504303, + "learning_rate": 1.6939039087447026e-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": 0.006136955763213337, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.002695623377803713, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.014969535078853369, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.01813545823097229, + "weighted_rkd_loss": 0.0074847675394266846 + }, + { + "epoch": 4.0871021775544385, + "flow_matching_loss": 0.018189919274300338, + "grad_norm": 0.19047412276268005, + "learning_rate": 1.6840621609248397e-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": 0.006888562859967351, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0030031787464395165, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.016780304722487926, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.018189919274300338, + "weighted_rkd_loss": 0.008390152361243963 + }, + { + "epoch": 4.1038525963149075, + "flow_matching_loss": 0.01785136330872774, + "grad_norm": 0.27839913964271545, + "learning_rate": 1.6742123616297794e-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": 0.006834283005446196, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0029112873249687254, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.016579853277653454, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.01785136330872774, + "weighted_rkd_loss": 0.008289926638826727 + }, + { + "epoch": 4.1206030150753765, + "flow_matching_loss": 0.016325243934988975, + "grad_norm": 0.15193955600261688, + "learning_rate": 1.664354941721651e-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": 0.00687442640773952, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.002769691066350788, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.016518543753772975, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.016325243934988975, + "weighted_rkd_loss": 0.008259271876886487 + }, + { + "epoch": 4.1373534338358455, + "flow_matching_loss": 0.016034547705203295, + "grad_norm": 0.20900852978229523, + "learning_rate": 1.6544903323959355e-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": 0.006941587897017598, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0030604396713897588, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.016943615395575762, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.016034547705203295, + "weighted_rkd_loss": 0.008471807697787881 + }, + { + "epoch": 4.1541038525963145, + "flow_matching_loss": 0.018592958617955447, + "grad_norm": 0.19960598647594452, + "learning_rate": 1.6446189651626023e-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": 0.006952088605612516, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.002908537397161126, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01681271456182003, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.018592958617955447, + "weighted_rkd_loss": 0.008406357280910015 + }, + { + "epoch": 4.1708542713567835, + "flow_matching_loss": 0.01994622778147459, + "grad_norm": 0.19966141879558563, + "learning_rate": 1.6347412718272316e-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": 0.0064809655770659445, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.00286123298574239, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.0158231639303267, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.01994622778147459, + "weighted_rkd_loss": 0.00791158196516335 + }, + { + "epoch": 4.187604690117253, + "flow_matching_loss": 0.015308264829218387, + "grad_norm": 0.20533037185668945, + "learning_rate": 1.624857684472131e-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": 0.006646136892959476, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0029631488490849734, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.016255422495305538, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.015308264829218387, + "weighted_rkd_loss": 0.008127711247652769 + }, + { + "epoch": 4.204355108877722, + "flow_matching_loss": 0.01632122229784727, + "grad_norm": 0.20436812937259674, + "learning_rate": 1.6149686354374283e-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": 0.006125012668780983, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.002678025362547487, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.014928050618618727, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.01632122229784727, + "weighted_rkd_loss": 0.007464025309309364 + }, + { + "epoch": 4.221105527638191, + "flow_matching_loss": 0.014456022158265113, + "grad_norm": 0.20241229236125946, + "learning_rate": 1.605074557302165e-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": 0.007364161545410752, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0034445435972884297, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01817286675795913, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.014456022158265113, + "weighted_rkd_loss": 0.009086433378979564 + }, + { + "epoch": 4.23785594639866, + "flow_matching_loss": 0.019517047703266142, + "grad_norm": 0.15849557518959045, + "learning_rate": 1.59517588286537e-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": 0.006639987695962191, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.002627583674620837, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01590755898505449, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.019517047703266142, + "weighted_rkd_loss": 0.007953779492527246 + }, + { + "epoch": 4.254606365159129, + "flow_matching_loss": 0.01830279529094696, + "grad_norm": 0.24347402155399323, + "learning_rate": 1.5852730451271324e-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": 0.007150422735139728, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.003178823774214834, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01747966939583421, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.01830279529094696, + "weighted_rkd_loss": 0.008739834697917104 + }, + { + "epoch": 4.271356783919598, + "flow_matching_loss": 0.01607086043804884, + "grad_norm": 0.16499897837638855, + "learning_rate": 1.5753664772696546e-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": 0.005285810353234411, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0025143951759673657, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.013086015731096268, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.01607086043804884, + "weighted_rkd_loss": 0.006543007865548134 + }, + { + "epoch": 4.288107202680067, + "flow_matching_loss": 0.017657784558832646, + "grad_norm": 0.19686786830425262, + "learning_rate": 1.565456612638307e-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": 0.006788184400647879, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0029420934384688734, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01651846235617995, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.017657784558832646, + "weighted_rkd_loss": 0.008259231178089976 + }, + { + "epoch": 4.304857621440536, + "flow_matching_loss": 0.017839622404426336, + "grad_norm": 0.15110993385314941, + "learning_rate": 1.5555438847226732e-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": 0.005660876887850464, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0021738476702012123, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.013495601434260606, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.017839622404426336, + "weighted_rkd_loss": 0.006747800717130303 + }, + { + "epoch": 4.321608040201005, + "flow_matching_loss": 0.018390034884214403, + "grad_norm": 0.19412873685359955, + "learning_rate": 1.545628727137584e-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": 0.006427357415668666, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0028176769730634986, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.015672391373664142, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.018390034884214403, + "weighted_rkd_loss": 0.007836195686832071 + }, + { + "epoch": 4.338358458961474, + "flow_matching_loss": 0.01622212715446949, + "grad_norm": 0.183297798037529, + "learning_rate": 1.535711573604153e-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": 0.007544292882084846, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0035195130738429727, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.018608098663389683, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.01622212715446949, + "weighted_rkd_loss": 0.009304049331694842 + }, + { + "epoch": 4.355108877721943, + "flow_matching_loss": 0.017579875886440277, + "grad_norm": 0.15577158331871033, + "learning_rate": 1.5257928579308032e-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": 0.00629424755461514, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.00317120929248631, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01575970444828272, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.017579875886440277, + "weighted_rkd_loss": 0.00787985222414136 + }, + { + "epoch": 4.371859296482412, + "flow_matching_loss": 0.017358833644539117, + "grad_norm": 0.22093512117862701, + "learning_rate": 1.5158730139942902e-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": 0.008002762030810118, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0033222919795662166, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01932781608775258, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.017358833644539117, + "weighted_rkd_loss": 0.00966390804387629 + }, + { + "epoch": 4.388609715242881, + "flow_matching_loss": 0.01748079704120755, + "grad_norm": 0.17871694266796112, + "learning_rate": 1.5059524757207239e-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": 0.006335989758372307, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.002982900128699839, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.015654879808425903, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.01748079704120755, + "weighted_rkd_loss": 0.007827439904212951 + }, + { + "epoch": 4.40536013400335, + "flow_matching_loss": 0.015273214224725962, + "grad_norm": 0.1625806838274002, + "learning_rate": 1.4960316770665866e-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": 0.006002801167778671, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.002634131000377238, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.0146397334523499, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.015273214224725962, + "weighted_rkd_loss": 0.00731986672617495 + }, + { + "epoch": 4.422110552763819, + "flow_matching_loss": 0.01621934836730361, + "grad_norm": 0.19268380105495453, + "learning_rate": 1.4861110519997495e-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": 0.0055179823189973835, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.002493456448428333, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.013529421016573905, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.01621934836730361, + "weighted_rkd_loss": 0.006764710508286953 + }, + { + "epoch": 4.438860971524288, + "flow_matching_loss": 0.016301322728395462, + "grad_norm": 0.220143124461174, + "learning_rate": 1.4761910344804913e-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": 0.006636293604969978, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0027302491129375995, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.016002836357802152, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.016301322728395462, + "weighted_rkd_loss": 0.008001418178901076 + }, + { + "epoch": 4.455611390284757, + "flow_matching_loss": 0.01781105054542422, + "grad_norm": 0.18916308879852295, + "learning_rate": 1.466272058442516e-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": 0.006512451078742742, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0026996355154551566, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.015724537800997494, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.01781105054542422, + "weighted_rkd_loss": 0.007862268900498747 + }, + { + "epoch": 4.472361809045226, + "flow_matching_loss": 0.01598687143996358, + "grad_norm": 0.20439769327640533, + "learning_rate": 1.4563545577739676e-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": 0.007010463485494256, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.003305570350494236, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.017326497286558152, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.01598687143996358, + "weighted_rkd_loss": 0.008663248643279076 + }, + { + "epoch": 4.489112227805695, + "flow_matching_loss": 0.015458551421761512, + "grad_norm": 0.2059379369020462, + "learning_rate": 1.4464389662984543e-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": 0.006447748886421323, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0029736610711552204, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01586915897205472, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.015458551421761512, + "weighted_rkd_loss": 0.00793457948602736 + }, + { + "epoch": 4.505862646566165, + "flow_matching_loss": 0.014416098967194558, + "grad_norm": 0.12666106224060059, + "learning_rate": 1.4365257177560684e-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": 0.005249654781073332, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0021719597512856125, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.012671269290149213, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.014416098967194558, + "weighted_rkd_loss": 0.006335634645074606 + }, + { + "epoch": 4.522613065326633, + "flow_matching_loss": 0.018572003673762084, + "grad_norm": 0.1562814861536026, + "learning_rate": 1.426615245784416e-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": 0.006474327272735536, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.00239322081906721, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.015341875329613685, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.018572003673762084, + "weighted_rkd_loss": 0.0076709376648068425 + }, + { + "epoch": 4.539363484087103, + "flow_matching_loss": 0.015323669649660587, + "grad_norm": 0.21075128018856049, + "learning_rate": 1.4167079838996468e-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": 0.006760964635759592, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.003025453584268689, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01654738299548626, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.015323669649660587, + "weighted_rkd_loss": 0.00827369149774313 + }, + { + "epoch": 4.556113902847571, + "flow_matching_loss": 0.016095327865332364, + "grad_norm": 0.18271535634994507, + "learning_rate": 1.4068043654774897e-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": 0.006994287949055433, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0034679821808822455, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01745655816048384, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.016095327865332364, + "weighted_rkd_loss": 0.00872827908024192 + }, + { + "epoch": 4.572864321608041, + "flow_matching_loss": 0.015398699510842561, + "grad_norm": 0.17756566405296326, + "learning_rate": 1.3969048237342992e-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": 0.005852023372426629, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0022669992642477153, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01397104598581791, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.015398699510842561, + "weighted_rkd_loss": 0.006985522992908955 + }, + { + "epoch": 4.589614740368509, + "flow_matching_loss": 0.019043789617717265, + "grad_norm": 0.18888011574745178, + "learning_rate": 1.3870097917081011e-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": 0.006620995048433542, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0027569215861149134, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01599891185760498, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.019043789617717265, + "weighted_rkd_loss": 0.00799945592880249 + }, + { + "epoch": 4.606365159128979, + "flow_matching_loss": 0.015605267789214849, + "grad_norm": 0.16311275959014893, + "learning_rate": 1.3771197022396508e-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": 0.005664682202041149, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.002713108283933252, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.014042472653090953, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.015605267789214849, + "weighted_rkd_loss": 0.007021236326545477 + }, + { + "epoch": 4.623115577889447, + "flow_matching_loss": 0.016418025083839893, + "grad_norm": 0.19125907123088837, + "learning_rate": 1.3672349879535015e-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": 0.0060530815739184615, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.002529566141311079, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.014635729324072599, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.016418025083839893, + "weighted_rkd_loss": 0.007317864662036299 + }, + { + "epoch": 4.639865996649917, + "flow_matching_loss": 0.01560895647853613, + "grad_norm": 0.21102729439735413, + "learning_rate": 1.3573560812390781e-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": 0.005735368002206087, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0025257673114538193, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01399650340899825, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.01560895647853613, + "weighted_rkd_loss": 0.006998251704499125 + }, + { + "epoch": 4.656616415410385, + "flow_matching_loss": 0.01519146803766489, + "grad_norm": 0.22826486825942993, + "learning_rate": 1.3474834142317635e-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": 0.007076088129542768, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.003330452204681933, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.017482628300786018, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.01519146803766489, + "weighted_rkd_loss": 0.008741314150393009 + }, + { + "epoch": 4.673366834170855, + "flow_matching_loss": 0.01835788758471608, + "grad_norm": 0.1848084032535553, + "learning_rate": 1.3376174187939955e-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": 0.0065226480830460785, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.002571828872896731, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.015617125015705825, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.01835788758471608, + "weighted_rkd_loss": 0.007808562507852912 + }, + { + "epoch": 4.690117252931323, + "flow_matching_loss": 0.018094559013843537, + "grad_norm": 0.190769225358963, + "learning_rate": 1.3277585264963761e-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": 0.00641796961426735, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.002544945338740945, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.015380884800106288, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.018094559013843537, + "weighted_rkd_loss": 0.007690442400053144 + }, + { + "epoch": 4.706867671691793, + "flow_matching_loss": 0.015870044007897377, + "grad_norm": 0.21869879961013794, + "learning_rate": 1.3179071685987928e-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": 0.006041633710265159, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0025358419166877865, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.014619109313935041, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.015870044007897377, + "weighted_rkd_loss": 0.0073095546569675205 + }, + { + "epoch": 4.723618090452261, + "flow_matching_loss": 0.016662956122308968, + "grad_norm": 0.18645018339157104, + "learning_rate": 1.308063776031554e-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": 0.006069857371039688, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.002291768416762352, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.014431483019143343, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.016662956122308968, + "weighted_rkd_loss": 0.007215741509571672 + }, + { + "epoch": 4.740368509212731, + "flow_matching_loss": 0.015010672435164452, + "grad_norm": 0.1860949844121933, + "learning_rate": 1.2982287793765389e-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": 0.005239035491831601, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0023728609085083008, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.012850932031869888, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.015010672435164452, + "weighted_rkd_loss": 0.006425466015934944 + }, + { + "epoch": 4.757118927973199, + "flow_matching_loss": 0.01519556799903512, + "grad_norm": 0.174799844622612, + "learning_rate": 1.2884026088483616e-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": 0.00624262671917677, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.002823116723448038, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.015308370254933834, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.01519556799903512, + "weighted_rkd_loss": 0.007654185127466917 + }, + { + "epoch": 4.773869346733669, + "flow_matching_loss": 0.01977842962369323, + "grad_norm": 0.1629028618335724, + "learning_rate": 1.2785856942755543e-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": 0.006113787461072206, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.003103880595881492, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.015331455413252116, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.01977842962369323, + "weighted_rkd_loss": 0.007665727706626058 + }, + { + "epoch": 4.790619765494137, + "flow_matching_loss": 0.0173368944786489, + "grad_norm": 0.20144276320934296, + "learning_rate": 1.2687784650817616e-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": 0.005899609881453216, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.00227838649880141, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.014077606238424778, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.0173368944786489, + "weighted_rkd_loss": 0.007038803119212389 + }, + { + "epoch": 4.807370184254607, + "flow_matching_loss": 0.018012631684541702, + "grad_norm": 0.20246005058288574, + "learning_rate": 1.2589813502669603e-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": 0.005113742733374238, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0021042732521891596, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.012331758718937635, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.018012631684541702, + "weighted_rkd_loss": 0.0061658793594688175 + }, + { + "epoch": 4.824120603015075, + "flow_matching_loss": 0.018170248344540597, + "grad_norm": 0.1560131311416626, + "learning_rate": 1.2491947783886907e-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": 0.006307589891366661, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0031363356858491896, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01575151551514864, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.018170248344540597, + "weighted_rkd_loss": 0.00787575775757432 + }, + { + "epoch": 4.840871021775545, + "flow_matching_loss": 0.01662774235010147, + "grad_norm": 0.159496009349823, + "learning_rate": 1.2394191775433106e-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": 0.005343643948435783, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0026379677816294133, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.013325255736708641, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.01662774235010147, + "weighted_rkd_loss": 0.006662627868354321 + }, + { + "epoch": 4.857621440536013, + "flow_matching_loss": 0.015159292798489332, + "grad_norm": 0.20265640318393707, + "learning_rate": 1.2296549753472696e-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": 0.005048206844367087, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0023854476399719714, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.012481861375272274, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.015159292798489332, + "weighted_rkd_loss": 0.006240930687636137 + }, + { + "epoch": 4.874371859296483, + "flow_matching_loss": 0.014173550438135863, + "grad_norm": 0.2068011462688446, + "learning_rate": 1.2199025989184032e-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": 0.005431518261320889, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0022641819319687783, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.013127218419685959, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.014173550438135863, + "weighted_rkd_loss": 0.0065636092098429796 + }, + { + "epoch": 4.891122278056951, + "flow_matching_loss": 0.01629276154562831, + "grad_norm": 0.14624083042144775, + "learning_rate": 1.21016247485725e-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": 0.005276934406720102, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0025586300413124262, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.013112498819828034, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.01629276154562831, + "weighted_rkd_loss": 0.006556249409914017 + }, + { + "epoch": 4.907872696817421, + "flow_matching_loss": 0.015676627214998008, + "grad_norm": 0.14908982813358307, + "learning_rate": 1.2004350292283897e-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": 0.005060716066509485, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0022387097356840967, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.012360141845420003, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.015676627214998008, + "weighted_rkd_loss": 0.0061800709227100015 + }, + { + "epoch": 4.924623115577889, + "flow_matching_loss": 0.015715810284018517, + "grad_norm": 0.16635891795158386, + "learning_rate": 1.1907206875418068e-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": 0.006101622781716287, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0026141874259337783, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.014817433059215545, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.015715810284018517, + "weighted_rkd_loss": 0.007408716529607773 + }, + { + "epoch": 4.941373534338359, + "flow_matching_loss": 0.01659667817875743, + "grad_norm": 0.1846851259469986, + "learning_rate": 1.1810198747342769e-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": 0.005292257177643478, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0024473055149428547, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.013031819835305214, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.01659667817875743, + "weighted_rkd_loss": 0.006515909917652607 + }, + { + "epoch": 4.958123953098827, + "flow_matching_loss": 0.013684543035924434, + "grad_norm": 0.15837141871452332, + "learning_rate": 1.1713330151507786e-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": 0.005822117114439607, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.002399850985966623, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.014044085331261159, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.013684543035924434, + "weighted_rkd_loss": 0.007022042665630579 + }, + { + "epoch": 4.974874371859297, + "flow_matching_loss": 0.01659283647313714, + "grad_norm": 0.17752093076705933, + "learning_rate": 1.161660532525931e-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": 0.0058163336478173734, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.002715875511057675, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.014348542969673871, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.01659283647313714, + "weighted_rkd_loss": 0.007174271484836936 + }, + { + "epoch": 4.991624790619765, + "flow_matching_loss": 0.014203921891748905, + "grad_norm": 0.19184257090091705, + "learning_rate": 1.1520028499654584e-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": 0.0061296496540308, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0023868637392297385, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.014646163024008275, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.014203921891748905, + "weighted_rkd_loss": 0.007323081512004137 + }, + { + "epoch": 5.008375209380235, + "flow_matching_loss": 0.014893015008419752, + "grad_norm": 0.2044752985239029, + "learning_rate": 1.1423603899276833e-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": 0.006520736729726195, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.002464273199439049, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.015505746752023698, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.014893015008419752, + "weighted_rkd_loss": 0.007752873376011849 + }, + { + "epoch": 5.025125628140704, + "flow_matching_loss": 0.015568806696683169, + "grad_norm": 0.22325921058654785, + "learning_rate": 1.1327335742050442e-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": 0.005686519085429609, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0026375182322226466, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.014010556414723397, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.015568806696683169, + "weighted_rkd_loss": 0.007005278207361698 + } + ], + "logging_steps": 10, + "max_steps": 5000, + "num_input_tokens_seen": 0, + "num_train_epochs": 9, + "save_steps": 5001, + "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": 64, + "trial_name": null, + "trial_params": null +} diff --git a/FewShot/39tasks/5shot/FromRKD-Timewarp_FullactionHead/seed1/default/train_5shot_seed1_fullactionhead_5k/experiment_cfg/metadata.json b/FewShot/39tasks/5shot/FromRKD-Timewarp_FullactionHead/seed1/default/train_5shot_seed1_fullactionhead_5k/experiment_cfg/metadata.json new file mode 100644 index 0000000000000000000000000000000000000000..27700cac5d292d4d42897b3c2e80bdc58e55c60e --- /dev/null +++ b/FewShot/39tasks/5shot/FromRKD-Timewarp_FullactionHead/seed1/default/train_5shot_seed1_fullactionhead_5k/experiment_cfg/metadata.json @@ -0,0 +1,431 @@ +{ + "new_embodiment": { + "statistics": { + "state": { + "base_position": { + "max": [ + 7.778976917266846, + 0.33403685688972473, + 0.7223060727119446 + ], + "min": [ + -4.821934223175049, + -6.785110950469971, + 0.6994727253913879 + ], + "mean": [ + 2.5719731235192254, + -2.1489371002267235, + 0.7003612741268546 + ], + "std": [ + 1.669945884010784, + 1.3739411381361726, + 0.0009413282367754987 + ], + "q01": [ + -0.7944286339399775, + -5.486611518471242, + 0.699977555971739 + ], + "q99": [ + 6.019886905263648, + -0.5229184210026855, + 0.702930326282967 + ] + }, + "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, + 0.248543484335273, + 0.5959317405459531 + ], + "std": [ + 0.0, + 0.0, + 0.6655550240074092, + 0.374283995143559 + ], + "q01": [ + 0.0, + 0.0, + -1.0000000000000002, + 4.503254968910557e-06 + ], + "q99": [ + 0.0, + 0.0, + 1.0000000000000002, + 1.0000000000000002 + ] + }, + "end_effector_position_relative": { + "max": [ + 0.8848381638526917, + 0.8569459915161133, + 1.053884744644165 + ], + "min": [ + -0.5828549861907959, + -0.8486236333847046, + -0.3446122705936432 + ], + "mean": [ + 0.2465490874130862, + -0.004711658908481467, + 0.43463835464312073 + ], + "std": [ + 0.15798784670671595, + 0.25624102400497073, + 0.20836028727974915 + ], + "q01": [ + -0.020752051134640608, + -0.4837933887347875, + 0.1802942952289915 + ], + "q99": [ + 0.5386784714844007, + 0.4801424212024721, + 0.695095732975766 + ] + }, + "end_effector_rotation_relative": { + "max": [ + 0.9999949932098389, + 0.9997354745864868, + 0.9900747537612915, + 0.9735639095306396 + ], + "min": [ + -0.9999914169311523, + -0.9999884366989136, + -0.9975369572639465, + 6.754255821306288e-08 + ], + "mean": [ + -0.22531859714056926, + 0.046973702821026614, + -0.0568176967878771, + 0.12364307042955697 + ], + "std": [ + 0.8502144333612052, + 0.29591250318147405, + 0.30218053156837016, + 0.1636031353412931 + ], + "q01": [ + -0.9959057387172566, + -0.5665384745573048, + -0.5329334620834977, + 0.0014372279999397553 + ], + "q99": [ + 0.9904107951565839, + 0.6359623612021396, + 0.5026076022439377, + 0.42955559234879914 + ] + }, + "gripper_qpos": { + "max": [ + 0.05289580300450325, + 0.00937709677964449 + ], + "min": [ + -0.011669117026031017, + -0.06307756155729294 + ], + "mean": [ + 0.03188757060429838, + -0.0321793333086053 + ], + "std": [ + 0.01171052313737148, + 0.011427336801203189 + ], + "q01": [ + 0.006462901319459925, + -0.040557140599419166 + ], + "q99": [ + 0.04037619335388386, + -0.006523947835993347 + ] + } + }, + "action": { + "base_motion": { + "max": [ + 1.0, + 1.0, + 1.0, + 0.0 + ], + "min": [ + -1.0, + -1.0, + -1.0, + 0.0 + ], + "mean": [ + 0.006639007355857881, + 0.0027330421795561874, + -0.0017774441970214892, + 0.0 + ], + "std": [ + 0.11841528114040736, + 0.1041359083946544, + 0.08346265521208351, + 0.0 + ], + "q01": [ + -0.17562359675270237, + -0.21706402321529558, + -0.16273392585723104, + 0.0 + ], + "q99": [ + 0.2572918701644887, + 0.27022325575787115, + 0.1369626751049278, + 0.0 + ] + }, + "control_mode": { + "max": [ + 1.0 + ], + "min": [ + -1.0 + ], + "mean": [ + -0.9205539608927732 + ], + "std": [ + 0.39063657242534666 + ], + "q01": [ + -1.0 + ], + "q99": [ + -0.17711863866562935 + ] + }, + "end_effector_position": { + "max": [ + 1.0, + 1.0, + 1.0 + ], + "min": [ + -1.0, + -1.0, + -1.0 + ], + "mean": [ + -0.00924803964621179, + 0.004053841761390575, + -0.08385115578460103 + ], + "std": [ + 0.45413072518223824, + 0.40158612714634856, + 0.41068634292849304 + ], + "q01": [ + -0.9669817145875343, + -0.9241307799783718, + -0.8264079503068825 + ], + "q99": [ + 0.8021511975722166, + 0.9241346582748478, + 0.8348116781502298 + ] + }, + "end_effector_rotation": { + "max": [ + 1.0, + 1.0, + 1.0 + ], + "min": [ + -1.0, + -1.0, + -1.0 + ], + "mean": [ + 0.007644157504353809, + -0.019962543093886335, + -0.002047110452602617 + ], + "std": [ + 0.11076012154791087, + 0.13176735140285795, + 0.12286528665895614 + ], + "q01": [ + -0.2659105439401192, + -0.3961694033813533, + -0.3434870832468936 + ], + "q99": [ + 0.3145786959266369, + 0.31309430361227547, + 0.34107754401942986 + ] + }, + "gripper_close": { + "max": [ + 1.0 + ], + "min": [ + -1.0 + ], + "mean": [ + -0.29870612267533575 + ], + "std": [ + 0.9543487555375691 + ], + "q01": [ + -1.0 + ], + "q99": [ + 0.7610674802484976 + ] + } + } + }, + "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/FewShot/39tasks/5shot/FromRKD-Timewarp_FullactionHead/seed2/default/resolved_config.yaml b/FewShot/39tasks/5shot/FromRKD-Timewarp_FullactionHead/seed2/default/resolved_config.yaml new file mode 100644 index 0000000000000000000000000000000000000000..ce9b1070764d237724a6133ede4cdcb191eaee27 --- /dev/null +++ b/FewShot/39tasks/5shot/FromRKD-Timewarp_FullactionHead/seed2/default/resolved_config.yaml @@ -0,0 +1,38 @@ +experiment_name: fewshot_39tasks_5shot_from_rkd_timewarp_fullactionhead_seed2 +policy_type: groot_RKDv3 +base_model_path: ${HOME}/groot_robocasa/robocasa_v2/TimewarpVAE/RKDv3_DA_APHead_from_Atomic26Baseline60k/default/phase3 +output_root: ${HOME}/groot_robocasa/robocasa_v2/FewShot/39tasks/5shot/FromRKD-Timewarp_FullactionHead/seed2 +dataset: + dataset_soup: fewshot_heldout_atomic39_human_5shot +training: + num_gpus: 1 + batch_size: 64 + seed: 42 + filter_key_seed: 2 + manifest_path: ${HOME}/clvla/benchmarks/robocasa_v2/my_scripts/Fewshot/manifests/fewshot_heldout_atomic39/shot5_seed2.json + save_total_limit: 4 +model: + rkd_enabled: true + rkd_fm_loss_weight: 1.0 + rkd_loss_weight: 0.5 + rkd_loss_weight_schedule: null +phases: +- name: train_5shot_seed2_fullactionhead_5k + max_steps: 5000 + save_steps: 0 + save_step_list: + - 1000 + - 3000 + - 5000 + save_final_model: false + 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: 0.5 + rkd_loss_weight_schedule: null +resolved_sweep: {} diff --git a/FewShot/39tasks/5shot/FromRKD-Timewarp_FullactionHead/seed2/default/seed2.yaml b/FewShot/39tasks/5shot/FromRKD-Timewarp_FullactionHead/seed2/default/seed2.yaml new file mode 100644 index 0000000000000000000000000000000000000000..5c6f35679cfb2661e8adc4682bbeeb5fe72e710a --- /dev/null +++ b/FewShot/39tasks/5shot/FromRKD-Timewarp_FullactionHead/seed2/default/seed2.yaml @@ -0,0 +1,37 @@ +experiment_name: fewshot_39tasks_5shot_from_rkd_timewarp_fullactionhead_seed2 +policy_type: groot_RKDv3 +base_model_path: ${HOME}/groot_robocasa/robocasa_v2/TimewarpVAE/RKDv3_DA_APHead_from_Atomic26Baseline60k/default/phase3 +output_root: ${HOME}/groot_robocasa/robocasa_v2/FewShot/39tasks/5shot/FromRKD-Timewarp_FullactionHead/seed2 +dataset: + dataset_soup: fewshot_heldout_atomic39_human_5shot +training: + num_gpus: 1 + batch_size: 64 + seed: 42 + filter_key_seed: 2 + manifest_path: ${HOME}/clvla/benchmarks/robocasa_v2/my_scripts/Fewshot/manifests/fewshot_heldout_atomic39/shot5_seed2.json + save_total_limit: 4 +model: + rkd_enabled: true + rkd_fm_loss_weight: 1.0 + rkd_loss_weight: 0.5 + rkd_loss_weight_schedule: null +phases: +- name: train_5shot_seed2_fullactionhead_5k + max_steps: 5000 + save_steps: 0 + save_step_list: + - 1000 + - 3000 + - 5000 + save_final_model: false + 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: 0.5 + rkd_loss_weight_schedule: null diff --git a/FewShot/39tasks/5shot/FromRKD-Timewarp_FullactionHead/seed2/default/train_5shot_seed2_fullactionhead_5k/checkpoint-1000/config.json b/FewShot/39tasks/5shot/FromRKD-Timewarp_FullactionHead/seed2/default/train_5shot_seed2_fullactionhead_5k/checkpoint-1000/config.json new file mode 100644 index 0000000000000000000000000000000000000000..4740d0455bef31c469ee5dab6d4a47d64df61665 --- /dev/null +++ b/FewShot/39tasks/5shot/FromRKD-Timewarp_FullactionHead/seed2/default/train_5shot_seed2_fullactionhead_5k/checkpoint-1000/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": 0.5, + "rkd_loss_weight_end": 0.0, + "rkd_loss_weight_schedule": null, + "rkd_loss_weight_start": 0.1, + "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/FewShot/39tasks/5shot/FromRKD-Timewarp_FullactionHead/seed2/default/train_5shot_seed2_fullactionhead_5k/checkpoint-1000/experiment_cfg/metadata.json b/FewShot/39tasks/5shot/FromRKD-Timewarp_FullactionHead/seed2/default/train_5shot_seed2_fullactionhead_5k/checkpoint-1000/experiment_cfg/metadata.json new file mode 100644 index 0000000000000000000000000000000000000000..23dcfa871a88e56280f945357aef0296bd071711 --- /dev/null +++ b/FewShot/39tasks/5shot/FromRKD-Timewarp_FullactionHead/seed2/default/train_5shot_seed2_fullactionhead_5k/checkpoint-1000/experiment_cfg/metadata.json @@ -0,0 +1,431 @@ +{ + "new_embodiment": { + "statistics": { + "state": { + "base_position": { + "max": [ + 7.778976917266846, + 0.33403685688972473, + 0.7223060727119446 + ], + "min": [ + -4.821934223175049, + -6.785110950469971, + 0.6994727253913879 + ], + "mean": [ + 2.577927505176127, + -2.1525618086425604, + 0.7003541659101092 + ], + "std": [ + 1.6720365948940972, + 1.3768414767482617, + 0.0009102921174316669 + ], + "q01": [ + -0.7605057497562686, + -5.518671947805291, + 0.6999764703640262 + ], + "q99": [ + 6.038879433964306, + -0.5168299981147954, + 0.7029182461269848 + ] + }, + "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, + 0.24782371371751144, + 0.5971093380149053 + ], + "std": [ + 0.0, + 0.0, + 0.6642559499983499, + 0.3751891785184018 + ], + "q01": [ + 0.0, + 0.0, + -1.0, + 4.194193576416142e-06 + ], + "q99": [ + 0.0, + 0.0, + 1.0, + 1.0 + ] + }, + "end_effector_position_relative": { + "max": [ + 0.8848381638526917, + 0.8569459915161133, + 1.053884744644165 + ], + "min": [ + -0.5828549861907959, + -0.8486236333847046, + -0.3446122705936432 + ], + "mean": [ + 0.25006971579930243, + -0.005716176032514759, + 0.4426027721228362 + ], + "std": [ + 0.15838384751435522, + 0.25141504139800513, + 0.20646816234620438 + ], + "q01": [ + -0.017003721077380903, + -0.47738320165276543, + 0.1917563409753314 + ], + "q99": [ + 0.5401029199137674, + 0.4706646609268599, + 0.6965586537529516 + ] + }, + "end_effector_rotation_relative": { + "max": [ + 0.9999949932098389, + 0.9997354745864868, + 0.9900747537612915, + 0.9735639095306396 + ], + "min": [ + -0.9999914169311523, + -0.9999884366989136, + -0.9975369572639465, + 6.754255821306288e-08 + ], + "mean": [ + -0.22531628862537245, + 0.045669797538445966, + -0.06098055586474368, + 0.12292005493282424 + ], + "std": [ + 0.8506341437448487, + 0.2880763358322842, + 0.30816573781739026, + 0.16366165154486362 + ], + "q01": [ + -0.9957510593732032, + -0.5584741016891311, + -0.5371647381159943, + 0.0013784815838340492 + ], + "q99": [ + 0.9899375209218226, + 0.6327878884078038, + 0.5071539024531994, + 0.43555975763914345 + ] + }, + "gripper_qpos": { + "max": [ + 0.05289580300450325, + 0.00937709677964449 + ], + "min": [ + -0.011669117026031017, + -0.06307756155729294 + ], + "mean": [ + 0.03196506153371015, + -0.03229791582989581 + ], + "std": [ + 0.01171640503534049, + 0.01140016755425199 + ], + "q01": [ + 0.006290779996034005, + -0.04059717954551848 + ], + "q99": [ + 0.04038165838477291, + -0.006363076424932449 + ] + } + }, + "action": { + "base_motion": { + "max": [ + 1.0, + 1.0, + 1.0, + 0.0 + ], + "min": [ + -1.0, + -1.0, + -1.0, + 0.0 + ], + "mean": [ + 0.006112990930475949, + 0.002504955933349057, + -0.001639409958610569, + 0.0 + ], + "std": [ + 0.11478077265783658, + 0.1002662368589144, + 0.0807258815139021, + 0.0 + ], + "q01": [ + -0.16969970921613756, + -0.19919537649977875, + -0.15012139866371593, + 0.0 + ], + "q99": [ + 0.2466625453958827, + 0.25040842772679267, + 0.12777701094549537, + 0.0 + ] + }, + "control_mode": { + "max": [ + 1.0 + ], + "min": [ + -1.0 + ], + "mean": [ + -0.9253067479599785 + ], + "std": [ + 0.3792397014797737 + ], + "q01": [ + -1.0 + ], + "q99": [ + -0.1974093068624371 + ] + }, + "end_effector_position": { + "max": [ + 1.0, + 1.0, + 1.0 + ], + "min": [ + -1.0, + -1.0, + -1.0 + ], + "mean": [ + -0.006754222051360652, + 0.0027854853184013867, + -0.08087572130330477 + ], + "std": [ + 0.45779905420789546, + 0.40578975219522756, + 0.4062694098132682 + ], + "q01": [ + -0.968295960290036, + -0.9240402337368848, + -0.8249359802202183 + ], + "q99": [ + 0.8046634915851651, + 0.9212243025654695, + 0.838433711437593 + ] + }, + "end_effector_rotation": { + "max": [ + 1.0, + 1.0, + 1.0 + ], + "min": [ + -1.0, + -1.0, + -1.0 + ], + "mean": [ + 0.008633749251857347, + -0.020406457524911718, + -0.0027639446362883006 + ], + "std": [ + 0.11393972872407104, + 0.13300186268009506, + 0.12192288302920105 + ], + "q01": [ + -0.2668579353776187, + -0.40299251206771697, + -0.34712957892172586 + ], + "q99": [ + 0.32748281264383927, + 0.3179028969292779, + 0.33459405510046736 + ] + }, + "gripper_close": { + "max": [ + 1.0 + ], + "min": [ + -1.0 + ], + "mean": [ + -0.3091635769453286 + ], + "std": [ + 0.9510122432876397 + ], + "q01": [ + -1.0 + ], + "q99": [ + 0.7643545615216344 + ] + } + } + }, + "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/FewShot/39tasks/5shot/FromRKD-Timewarp_FullactionHead/seed2/default/train_5shot_seed2_fullactionhead_5k/checkpoint-1000/model.safetensors.index.json b/FewShot/39tasks/5shot/FromRKD-Timewarp_FullactionHead/seed2/default/train_5shot_seed2_fullactionhead_5k/checkpoint-1000/model.safetensors.index.json new file mode 100644 index 0000000000000000000000000000000000000000..b20baed71f4d16a411223b70d45d934c6cbbfe97 --- /dev/null +++ b/FewShot/39tasks/5shot/FromRKD-Timewarp_FullactionHead/seed2/default/train_5shot_seed2_fullactionhead_5k/checkpoint-1000/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/FewShot/39tasks/5shot/FromRKD-Timewarp_FullactionHead/seed2/default/train_5shot_seed2_fullactionhead_5k/checkpoint-1000/trainer_state.json b/FewShot/39tasks/5shot/FromRKD-Timewarp_FullactionHead/seed2/default/train_5shot_seed2_fullactionhead_5k/checkpoint-1000/trainer_state.json new file mode 100644 index 0000000000000000000000000000000000000000..6abcd4891a315f54d8f5a245e8903b257408204b --- /dev/null +++ b/FewShot/39tasks/5shot/FromRKD-Timewarp_FullactionHead/seed2/default/train_5shot_seed2_fullactionhead_5k/checkpoint-1000/trainer_state.json @@ -0,0 +1,4634 @@ +{ + "best_global_step": null, + "best_metric": null, + "best_model_checkpoint": null, + "epoch": 1.7094017094017095, + "eval_steps": 500, + "global_step": 1000, + "is_hyper_param_search": false, + "is_local_process_zero": true, + "is_world_process_zero": true, + "log_history": [ + { + "epoch": 0.017094017094017096, + "flow_matching_loss": 0.1910901501774788, + "grad_norm": 1.1204875707626343, + "learning_rate": 1.08e-06, + "loss": 0.393, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.14386700689792634, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.11605296358466148, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.4037869781255722, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.1910901501774788, + "weighted_rkd_loss": 0.2018934890627861 + }, + { + "epoch": 0.03418803418803419, + "flow_matching_loss": 0.19379425793886185, + "grad_norm": 1.295363187789917, + "learning_rate": 2.28e-06, + "loss": 0.3834, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.1352515198290348, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.10873031765222549, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.3792333573102951, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.19379425793886185, + "weighted_rkd_loss": 0.18961667865514756 + }, + { + "epoch": 0.05128205128205128, + "flow_matching_loss": 0.17340936958789827, + "grad_norm": 0.9476425051689148, + "learning_rate": 3.48e-06, + "loss": 0.3566, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.1321360796689987, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.10214361175894737, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.36641576886177063, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.17340936958789827, + "weighted_rkd_loss": 0.18320788443088531 + }, + { + "epoch": 0.06837606837606838, + "flow_matching_loss": 0.16260478422045707, + "grad_norm": 0.8357043862342834, + "learning_rate": 4.68e-06, + "loss": 0.3417, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.12940022349357605, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.09945235326886177, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.3582527995109558, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.16260478422045707, + "weighted_rkd_loss": 0.1791263997554779 + }, + { + "epoch": 0.08547008547008547, + "flow_matching_loss": 0.13773404359817504, + "grad_norm": 0.7658578157424927, + "learning_rate": 5.8800000000000005e-06, + "loss": 0.3112, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.12533388957381247, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.09635578840970993, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.3470235675573349, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.13773404359817504, + "weighted_rkd_loss": 0.17351178377866744 + }, + { + "epoch": 0.10256410256410256, + "flow_matching_loss": 0.10675260052084923, + "grad_norm": 0.6615934371948242, + "learning_rate": 7.08e-06, + "loss": 0.2707, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.11958464086055756, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.08864015787839889, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.3278094381093979, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.10675260052084923, + "weighted_rkd_loss": 0.16390471905469894 + }, + { + "epoch": 0.11965811965811966, + "flow_matching_loss": 0.09638643190264702, + "grad_norm": 0.6668271422386169, + "learning_rate": 8.28e-06, + "loss": 0.2454, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.11045710295438767, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.07702444642782211, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.2979386538267136, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.09638643190264702, + "weighted_rkd_loss": 0.1489693269133568 + }, + { + "epoch": 0.13675213675213677, + "flow_matching_loss": 0.08751716017723084, + "grad_norm": 0.7303902506828308, + "learning_rate": 9.48e-06, + "loss": 0.2274, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.10436114817857742, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.07100482136011124, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.2797271177172661, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.08751716017723084, + "weighted_rkd_loss": 0.13986355885863305 + }, + { + "epoch": 0.15384615384615385, + "flow_matching_loss": 0.07772385030984878, + "grad_norm": 0.7202248573303223, + "learning_rate": 1.068e-05, + "loss": 0.2108, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.09959760904312134, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.06698438599705696, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.2661796033382416, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.07772385030984878, + "weighted_rkd_loss": 0.1330898016691208 + }, + { + "epoch": 0.17094017094017094, + "flow_matching_loss": 0.07342424914240837, + "grad_norm": 0.5669143199920654, + "learning_rate": 1.1880000000000001e-05, + "loss": 0.1941, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.09150870665907859, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.05835277736186981, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.24137018769979476, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.07342424914240837, + "weighted_rkd_loss": 0.12068509384989738 + }, + { + "epoch": 0.18803418803418803, + "flow_matching_loss": 0.07115557119250297, + "grad_norm": 0.5177780985832214, + "learning_rate": 1.308e-05, + "loss": 0.1825, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.08484098091721534, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.05309118740260601, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.22277315109968185, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.07115557119250297, + "weighted_rkd_loss": 0.11138657554984092 + }, + { + "epoch": 0.20512820512820512, + "flow_matching_loss": 0.06366660706698894, + "grad_norm": 0.6365015506744385, + "learning_rate": 1.428e-05, + "loss": 0.1748, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.08515534773468972, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0519100297242403, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.2222207233309746, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.06366660706698894, + "weighted_rkd_loss": 0.1111103616654873 + }, + { + "epoch": 0.2222222222222222, + "flow_matching_loss": 0.06294491589069366, + "grad_norm": 0.5627866983413696, + "learning_rate": 1.548e-05, + "loss": 0.165, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.07825233079493046, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.04756654389202595, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.20407120436429976, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.06294491589069366, + "weighted_rkd_loss": 0.10203560218214988 + }, + { + "epoch": 0.23931623931623933, + "flow_matching_loss": 0.061772428825497626, + "grad_norm": 0.6260324716567993, + "learning_rate": 1.6680000000000003e-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": 0.0717402745038271, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.04264889135956764, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.18612944036722184, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.061772428825497626, + "weighted_rkd_loss": 0.09306472018361092 + }, + { + "epoch": 0.2564102564102564, + "flow_matching_loss": 0.061649881303310394, + "grad_norm": 0.6016267538070679, + "learning_rate": 1.7879999999999998e-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": 0.06565237119793892, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.03857212290167809, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.16987686604261398, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.061649881303310394, + "weighted_rkd_loss": 0.08493843302130699 + }, + { + "epoch": 0.27350427350427353, + "flow_matching_loss": 0.06200782023370266, + "grad_norm": 0.5819275379180908, + "learning_rate": 1.908e-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": 0.06578411012887955, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.03858097940683365, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.17014919966459274, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.06200782023370266, + "weighted_rkd_loss": 0.08507459983229637 + }, + { + "epoch": 0.2905982905982906, + "flow_matching_loss": 0.05749040208756924, + "grad_norm": 0.537067711353302, + "learning_rate": 2.0280000000000002e-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": 0.06401401236653328, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.036967079155147076, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.16499510556459426, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.05749040208756924, + "weighted_rkd_loss": 0.08249755278229713 + }, + { + "epoch": 0.3076923076923077, + "flow_matching_loss": 0.05535002276301384, + "grad_norm": 0.4923454225063324, + "learning_rate": 2.148e-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": 0.05838514976203442, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.033644383773207664, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.1504146859049797, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.05535002276301384, + "weighted_rkd_loss": 0.07520734295248985 + }, + { + "epoch": 0.3247863247863248, + "flow_matching_loss": 0.05286722965538502, + "grad_norm": 0.48796457052230835, + "learning_rate": 2.268e-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": 0.05483606085181236, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.031941465102136135, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.1416135847568512, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.05286722965538502, + "weighted_rkd_loss": 0.0708067923784256 + }, + { + "epoch": 0.3418803418803419, + "flow_matching_loss": 0.05254113264381886, + "grad_norm": 0.4936208128929138, + "learning_rate": 2.3880000000000002e-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": 0.05183771774172783, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.029638106748461724, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.13331354185938835, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.05254113264381886, + "weighted_rkd_loss": 0.06665677092969417 + }, + { + "epoch": 0.358974358974359, + "flow_matching_loss": 0.04968957602977753, + "grad_norm": 0.5536408424377441, + "learning_rate": 2.508e-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": 0.047118809819221494, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.02699419390410185, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.12123181372880935, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.04968957602977753, + "weighted_rkd_loss": 0.06061590686440468 + }, + { + "epoch": 0.37606837606837606, + "flow_matching_loss": 0.04866243675351143, + "grad_norm": 0.48103460669517517, + "learning_rate": 2.628e-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": 0.046376798301935196, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.02550226878374815, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.11825586706399918, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.04866243675351143, + "weighted_rkd_loss": 0.05912793353199959 + }, + { + "epoch": 0.39316239316239315, + "flow_matching_loss": 0.04957115277647972, + "grad_norm": 0.5273743271827698, + "learning_rate": 2.748e-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": 0.04591682404279709, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.02485441230237484, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.11668805927038192, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.04957115277647972, + "weighted_rkd_loss": 0.05834402963519096 + }, + { + "epoch": 0.41025641025641024, + "flow_matching_loss": 0.04972734041512013, + "grad_norm": 0.48441460728645325, + "learning_rate": 2.868e-05, + "loss": 0.105, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.04358960241079331, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.023280760645866393, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.11045996621251106, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.04972734041512013, + "weighted_rkd_loss": 0.05522998310625553 + }, + { + "epoch": 0.42735042735042733, + "flow_matching_loss": 0.046684912964701654, + "grad_norm": 0.45080485939979553, + "learning_rate": 2.9880000000000002e-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": 0.04428564272820949, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.02520337775349617, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.1137746624648571, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.046684912964701654, + "weighted_rkd_loss": 0.05688733123242855 + }, + { + "epoch": 0.4444444444444444, + "flow_matching_loss": 0.04779224582016468, + "grad_norm": 0.43459874391555786, + "learning_rate": 2.9999734259635676e-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": 0.036030808463692665, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.019671429321169852, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.0917330451309681, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.04779224582016468, + "weighted_rkd_loss": 0.04586652256548405 + }, + { + "epoch": 0.46153846153846156, + "flow_matching_loss": 0.04915213659405708, + "grad_norm": 0.4534752368927002, + "learning_rate": 2.9998815663057245e-05, + "loss": 0.0983, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.039135459996759894, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.01992818871513009, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.09819910898804665, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.04915213659405708, + "weighted_rkd_loss": 0.049099554494023326 + }, + { + "epoch": 0.47863247863247865, + "flow_matching_loss": 0.044950670748949054, + "grad_norm": 0.4854293167591095, + "learning_rate": 2.999724096969228e-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": 0.03963045179843903, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.02123024109750986, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.10049114525318145, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.044950670748949054, + "weighted_rkd_loss": 0.050245572626590726 + }, + { + "epoch": 0.49572649572649574, + "flow_matching_loss": 0.045206136628985406, + "grad_norm": 0.4032607972621918, + "learning_rate": 2.9995010248422984e-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": 0.03566844779998064, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.018644866067916155, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.08998176082968712, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.045206136628985406, + "weighted_rkd_loss": 0.04499088041484356 + }, + { + "epoch": 0.5128205128205128, + "flow_matching_loss": 0.042293429002165796, + "grad_norm": 0.4287409782409668, + "learning_rate": 2.999212359682832e-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": 0.03817300200462341, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.020988442189991473, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.09733444601297378, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.042293429002165796, + "weighted_rkd_loss": 0.04866722300648689 + }, + { + "epoch": 0.5299145299145299, + "flow_matching_loss": 0.04139162935316563, + "grad_norm": 0.45648202300071716, + "learning_rate": 2.9988581141179787e-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": 0.03135862443596125, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.016660952381789683, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.07937820106744767, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.04139162935316563, + "weighted_rkd_loss": 0.039689100533723834 + }, + { + "epoch": 0.5470085470085471, + "flow_matching_loss": 0.041223404183983806, + "grad_norm": 0.4205642342567444, + "learning_rate": 2.998438303643587e-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": 0.034546280093491075, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.018318970408290624, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.08741152957081795, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.041223404183983806, + "weighted_rkd_loss": 0.043705764785408975 + }, + { + "epoch": 0.5641025641025641, + "flow_matching_loss": 0.04060795344412327, + "grad_norm": 0.402844101190567, + "learning_rate": 2.997952946623527e-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": 0.0335417652502656, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.018018722627311946, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.08510225303471089, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.04060795344412327, + "weighted_rkd_loss": 0.042551126517355445 + }, + { + "epoch": 0.5811965811965812, + "flow_matching_loss": 0.04365103021264076, + "grad_norm": 0.37454837560653687, + "learning_rate": 2.9974020642888875e-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": 0.03194001261144876, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.016204175911843775, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.08008420169353485, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.04365103021264076, + "weighted_rkd_loss": 0.040042100846767424 + }, + { + "epoch": 0.5982905982905983, + "flow_matching_loss": 0.039248546957969664, + "grad_norm": 0.38750800490379333, + "learning_rate": 2.9967856807370458e-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": 0.03014408703893423, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.015359362401068211, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.07564753592014313, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.039248546957969664, + "weighted_rkd_loss": 0.037823767960071565 + }, + { + "epoch": 0.6153846153846154, + "flow_matching_loss": 0.04174409061670303, + "grad_norm": 0.4128570556640625, + "learning_rate": 2.996103822930615e-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": 0.030098697915673257, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.015234342124313117, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.07543173767626285, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.04174409061670303, + "weighted_rkd_loss": 0.037715868838131425 + }, + { + "epoch": 0.6324786324786325, + "flow_matching_loss": 0.038203606754541396, + "grad_norm": 0.3553514778614044, + "learning_rate": 2.9953565206962653e-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": 0.03010321594774723, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.015880199056118725, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.07608662992715835, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.038203606754541396, + "weighted_rkd_loss": 0.038043314963579176 + }, + { + "epoch": 0.6495726495726496, + "flow_matching_loss": 0.038802698254585266, + "grad_norm": 0.393643319606781, + "learning_rate": 2.9945438067234172e-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": 0.02764769084751606, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.013171730842441321, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.06846711188554763, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.038802698254585266, + "weighted_rkd_loss": 0.034233555942773816 + }, + { + "epoch": 0.6666666666666666, + "flow_matching_loss": 0.03741585947573185, + "grad_norm": 0.38524219393730164, + "learning_rate": 2.9936657165628124e-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": 0.0301553001627326, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.015053990110754967, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.07536459155380726, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.03741585947573185, + "weighted_rkd_loss": 0.03768229577690363 + }, + { + "epoch": 0.6837606837606838, + "flow_matching_loss": 0.037859869003295896, + "grad_norm": 0.3669644594192505, + "learning_rate": 2.992722288624959e-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": 0.02703232318162918, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.014790859073400497, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.06885550543665886, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.037859869003295896, + "weighted_rkd_loss": 0.03442775271832943 + }, + { + "epoch": 0.7008547008547008, + "flow_matching_loss": 0.03818879183381796, + "grad_norm": 0.30247363448143005, + "learning_rate": 2.9917135641784515e-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": 0.026851335540413857, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.013741841074079275, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.06744451224803924, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.03818879183381796, + "weighted_rkd_loss": 0.03372225612401962 + }, + { + "epoch": 0.717948717948718, + "flow_matching_loss": 0.0374411903321743, + "grad_norm": 0.4039696753025055, + "learning_rate": 2.9906395873481655e-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": 0.024242410250008107, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.012942939531058073, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.061427760496735574, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.0374411903321743, + "weighted_rkd_loss": 0.030713880248367787 + }, + { + "epoch": 0.7350427350427351, + "flow_matching_loss": 0.0361711286008358, + "grad_norm": 0.34961435198783875, + "learning_rate": 2.989500405113326e-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": 0.024853001721203326, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.011823360808193684, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.061529364064335824, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.0361711286008358, + "weighted_rkd_loss": 0.030764682032167912 + }, + { + "epoch": 0.7521367521367521, + "flow_matching_loss": 0.03577544335275888, + "grad_norm": 0.3615719676017761, + "learning_rate": 2.9882960673054547e-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": 0.02502665389329195, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.01260580299422145, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.06265911161899566, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.03577544335275888, + "weighted_rkd_loss": 0.03132955580949783 + }, + { + "epoch": 0.7692307692307693, + "flow_matching_loss": 0.03441168405115604, + "grad_norm": 0.3467264175415039, + "learning_rate": 2.9870266266061886e-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": 0.02530351039022207, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.011560741625726224, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.062167763337492944, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.03441168405115604, + "weighted_rkd_loss": 0.031083881668746472 + }, + { + "epoch": 0.7863247863247863, + "flow_matching_loss": 0.03624687343835831, + "grad_norm": 0.3562852442264557, + "learning_rate": 2.9856921385449766e-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": 0.025862449035048484, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.012488656770437956, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.06421355418860912, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.03624687343835831, + "weighted_rkd_loss": 0.03210677709430456 + }, + { + "epoch": 0.8034188034188035, + "flow_matching_loss": 0.03741344213485718, + "grad_norm": 0.3594230115413666, + "learning_rate": 2.9842926614966502e-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": 0.021881727501749992, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.01110158571973443, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.054865040257573126, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.03741344213485718, + "weighted_rkd_loss": 0.027432520128786563 + }, + { + "epoch": 0.8205128205128205, + "flow_matching_loss": 0.029932724311947823, + "grad_norm": 0.3690202832221985, + "learning_rate": 2.982828256678868e-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": 0.02407331857830286, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.012713254615664481, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.060859891772270205, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.029932724311947823, + "weighted_rkd_loss": 0.030429945886135103 + }, + { + "epoch": 0.8376068376068376, + "flow_matching_loss": 0.03617270402610302, + "grad_norm": 0.38251274824142456, + "learning_rate": 2.9812989881494412e-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": 0.025536379404366018, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.012556855008006096, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.06362961418926716, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.03617270402610302, + "weighted_rkd_loss": 0.03181480709463358 + }, + { + "epoch": 0.8547008547008547, + "flow_matching_loss": 0.03683352377265692, + "grad_norm": 0.3576796352863312, + "learning_rate": 2.9797049228035296e-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": 0.02277378421276808, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.011028627213090659, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.056576195359230044, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.03683352377265692, + "weighted_rkd_loss": 0.028288097679615022 + }, + { + "epoch": 0.8717948717948718, + "flow_matching_loss": 0.034184624627232554, + "grad_norm": 0.4046909511089325, + "learning_rate": 2.978046130370715e-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": 0.02267690245062113, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.011717867571860552, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.05707167312502861, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.034184624627232554, + "weighted_rkd_loss": 0.028535836562514304 + }, + { + "epoch": 0.8888888888888888, + "flow_matching_loss": 0.03213700447231531, + "grad_norm": 0.3536110818386078, + "learning_rate": 2.9763226834119535e-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": 0.02117164060473442, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.01059307949617505, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.052936359867453577, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.03213700447231531, + "weighted_rkd_loss": 0.026468179933726788 + }, + { + "epoch": 0.905982905982906, + "flow_matching_loss": 0.03486904371529818, + "grad_norm": 0.3733022212982178, + "learning_rate": 2.9745346573163963e-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": 0.02185452114790678, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.01133495531976223, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.05504399687051773, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.03486904371529818, + "weighted_rkd_loss": 0.027521998435258866 + }, + { + "epoch": 0.9230769230769231, + "flow_matching_loss": 0.031505046598613265, + "grad_norm": 0.35847941040992737, + "learning_rate": 2.9726821302980987e-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": 0.020499915070831775, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.01087835356593132, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.05187818445265293, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.031505046598613265, + "weighted_rkd_loss": 0.025939092226326465 + }, + { + "epoch": 0.9401709401709402, + "flow_matching_loss": 0.030537511594593524, + "grad_norm": 0.3062575161457062, + "learning_rate": 2.9707651833925935e-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": 0.019180506095290185, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.009259850112721324, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.04762086160480976, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.030537511594593524, + "weighted_rkd_loss": 0.02381043080240488 + }, + { + "epoch": 0.9572649572649573, + "flow_matching_loss": 0.033962773904204366, + "grad_norm": 0.33466994762420654, + "learning_rate": 2.968783900453348e-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": 0.019049630872905253, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.009260836755856872, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.04736009836196899, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.033962773904204366, + "weighted_rkd_loss": 0.023680049180984496 + }, + { + "epoch": 0.9743589743589743, + "flow_matching_loss": 0.03441021610051394, + "grad_norm": 0.3329583704471588, + "learning_rate": 2.9667383681480968e-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": 0.02119819391518831, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.00991179826669395, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.052308186143636706, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.03441021610051394, + "weighted_rkd_loss": 0.026154093071818353 + }, + { + "epoch": 0.9914529914529915, + "flow_matching_loss": 0.032573466189205647, + "grad_norm": 0.3133791387081146, + "learning_rate": 2.9646286759550493e-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": 0.019760795310139655, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.00934667899273336, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.048868270218372346, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.032573466189205647, + "weighted_rkd_loss": 0.024434135109186173 + }, + { + "epoch": 1.0085470085470085, + "flow_matching_loss": 0.030896119587123395, + "grad_norm": 0.2655918002128601, + "learning_rate": 2.9624549161589764e-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": 0.018024975061416627, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.008287688577547669, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.04433763958513737, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.030896119587123395, + "weighted_rkd_loss": 0.022168819792568684 + }, + { + "epoch": 1.0256410256410255, + "flow_matching_loss": 0.02869145143777132, + "grad_norm": 0.2975919544696808, + "learning_rate": 2.9602171838471745e-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": 0.01704270886257291, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.008086427114903927, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.042171845026314256, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.02869145143777132, + "weighted_rkd_loss": 0.021085922513157128 + }, + { + "epoch": 1.0427350427350428, + "flow_matching_loss": 0.028883341886103153, + "grad_norm": 0.25782883167266846, + "learning_rate": 2.957915576905303e-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": 0.020831129886209963, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.01092052492313087, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.052582784369587895, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.028883341886103153, + "weighted_rkd_loss": 0.026291392184793948 + }, + { + "epoch": 1.0598290598290598, + "flow_matching_loss": 0.02891363874077797, + "grad_norm": 0.3048412501811981, + "learning_rate": 2.9555501960131058e-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": 0.018006673827767373, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.00887504410929978, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.044888391718268396, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.02891363874077797, + "weighted_rkd_loss": 0.022444195859134198 + }, + { + "epoch": 1.0769230769230769, + "flow_matching_loss": 0.029369746521115302, + "grad_norm": 0.3183818757534027, + "learning_rate": 2.9531211446400068e-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": 0.016618114057928325, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.007927177706733346, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.041163405403494835, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.029369746521115302, + "weighted_rkd_loss": 0.020581702701747417 + }, + { + "epoch": 1.0940170940170941, + "flow_matching_loss": 0.028417748771607874, + "grad_norm": 0.2978658378124237, + "learning_rate": 2.9506285290405813e-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": 0.01775301033630967, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.007838332653045654, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.04334435388445854, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.028417748771607874, + "weighted_rkd_loss": 0.02167217694222927 + }, + { + "epoch": 1.1111111111111112, + "flow_matching_loss": 0.030568437837064266, + "grad_norm": 0.27075809240341187, + "learning_rate": 2.948072458249911e-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": 0.01854327218607068, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.009335583634674549, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.04642212763428688, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.030568437837064266, + "weighted_rkd_loss": 0.02321106381714344 + }, + { + "epoch": 1.1282051282051282, + "flow_matching_loss": 0.03013926316052675, + "grad_norm": 0.31489095091819763, + "learning_rate": 2.9454530440788125e-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": 0.018000216316431762, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.008591329725459219, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.044591762498021124, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.03013926316052675, + "weighted_rkd_loss": 0.022295881249010562 + }, + { + "epoch": 1.1452991452991452, + "flow_matching_loss": 0.03240344282239675, + "grad_norm": 0.3196583092212677, + "learning_rate": 2.9427704011089474e-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": 0.01813620710745454, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.008372926386073232, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.044645340740680696, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.03240344282239675, + "weighted_rkd_loss": 0.022322670370340348 + }, + { + "epoch": 1.1623931623931625, + "flow_matching_loss": 0.02988831289112568, + "grad_norm": 0.33447059988975525, + "learning_rate": 2.94002464668781e-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": 0.017554160580039023, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.008155462937429548, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.04326378367841244, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.02988831289112568, + "weighted_rkd_loss": 0.02163189183920622 + }, + { + "epoch": 1.1794871794871795, + "flow_matching_loss": 0.0297787019982934, + "grad_norm": 0.25146859884262085, + "learning_rate": 2.9372159009235936e-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": 0.016257153637707234, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.007763447891920805, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.040277755446732046, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.0297787019982934, + "weighted_rkd_loss": 0.020138877723366023 + }, + { + "epoch": 1.1965811965811965, + "flow_matching_loss": 0.033523568883538246, + "grad_norm": 0.2588149905204773, + "learning_rate": 2.9343442866799365e-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": 0.016528852004557847, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.007534453133121133, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.0405921570956707, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.033523568883538246, + "weighted_rkd_loss": 0.02029607854783535 + }, + { + "epoch": 1.2136752136752136, + "flow_matching_loss": 0.029541657492518424, + "grad_norm": 0.2806798219680786, + "learning_rate": 2.9314099295705495e-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": 0.017626770399510862, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.008789777662605048, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.04404331874102354, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.029541657492518424, + "weighted_rkd_loss": 0.02202165937051177 + }, + { + "epoch": 1.2307692307692308, + "flow_matching_loss": 0.02779527623206377, + "grad_norm": 0.23124118149280548, + "learning_rate": 2.928412957953718e-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": 0.017632605507969856, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.008616828918457031, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.043882039934396745, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.02779527623206377, + "weighted_rkd_loss": 0.021941019967198373 + }, + { + "epoch": 1.2478632478632479, + "flow_matching_loss": 0.028330446407198905, + "grad_norm": 0.24973063170909882, + "learning_rate": 2.925353502926689e-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": 0.015281323064118623, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.007210866408422589, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.037773511931300165, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.028330446407198905, + "weighted_rkd_loss": 0.018886755965650082 + }, + { + "epoch": 1.264957264957265, + "flow_matching_loss": 0.026782910339534283, + "grad_norm": 0.29868558049201965, + "learning_rate": 2.922231698319937e-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": 0.016518175043165682, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.008065681951120496, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.04110203199088573, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.026782910339534283, + "weighted_rkd_loss": 0.020551015995442866 + }, + { + "epoch": 1.282051282051282, + "flow_matching_loss": 0.026428927294909955, + "grad_norm": 0.2616300582885742, + "learning_rate": 2.9190476806913088e-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": 0.015521802846342326, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.007240159553475678, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.03828376531600952, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.026428927294909955, + "weighted_rkd_loss": 0.01914188265800476 + }, + { + "epoch": 1.2991452991452992, + "flow_matching_loss": 0.02718909680843353, + "grad_norm": 0.2325822114944458, + "learning_rate": 2.9158015893200504e-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": 0.016340584587305786, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.006903730751946569, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.03958490006625652, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.02718909680843353, + "weighted_rkd_loss": 0.01979245003312826 + }, + { + "epoch": 1.3162393162393162, + "flow_matching_loss": 0.028375131264328956, + "grad_norm": 0.32508528232574463, + "learning_rate": 2.9124935662007137e-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": 0.015554658975452185, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.007059313822537661, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.03816863112151623, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.028375131264328956, + "weighted_rkd_loss": 0.019084315560758115 + }, + { + "epoch": 1.3333333333333333, + "flow_matching_loss": 0.027656448632478715, + "grad_norm": 0.27773064374923706, + "learning_rate": 2.9091237560369467e-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": 0.01565126357600093, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.007432594196870923, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.03873512130230665, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.027656448632478715, + "weighted_rkd_loss": 0.019367560651153325 + }, + { + "epoch": 1.3504273504273505, + "flow_matching_loss": 0.02753492761403322, + "grad_norm": 0.2521279454231262, + "learning_rate": 2.905692306235163e-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": 0.014433691278100014, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0070426968391984705, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.03591007981449366, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.02753492761403322, + "weighted_rkd_loss": 0.01795503990724683 + }, + { + "epoch": 1.3675213675213675, + "flow_matching_loss": 0.027198122441768648, + "grad_norm": 0.2660417854785919, + "learning_rate": 2.9021993668980925e-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": 0.015085380058735609, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.007494675228372216, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.03766543548554182, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.027198122441768648, + "weighted_rkd_loss": 0.01883271774277091 + }, + { + "epoch": 1.3846153846153846, + "flow_matching_loss": 0.031527518108487126, + "grad_norm": 0.3124120831489563, + "learning_rate": 2.8986450908182185e-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": 0.015878252685070038, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.007188629452139139, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.038945135287940504, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.031527518108487126, + "weighted_rkd_loss": 0.019472567643970252 + }, + { + "epoch": 1.4017094017094016, + "flow_matching_loss": 0.028070161677896978, + "grad_norm": 0.23588840663433075, + "learning_rate": 2.8950296334710902e-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": 0.016684133745729923, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.007126115215942264, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.040494382940232755, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.028070161677896978, + "weighted_rkd_loss": 0.020247191470116378 + }, + { + "epoch": 1.4188034188034189, + "flow_matching_loss": 0.02908680960536003, + "grad_norm": 0.24586918950080872, + "learning_rate": 2.8913531530085248e-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": 0.014786572381854057, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.007374308165162801, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.036947453022003175, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.02908680960536003, + "weighted_rkd_loss": 0.018473726511001588 + }, + { + "epoch": 1.435897435897436, + "flow_matching_loss": 0.026878987438976765, + "grad_norm": 0.22994333505630493, + "learning_rate": 2.887615810251687e-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": 0.01679125111550093, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.007170996256172657, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.040753497555851934, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.026878987438976765, + "weighted_rkd_loss": 0.020376748777925967 + }, + { + "epoch": 1.452991452991453, + "flow_matching_loss": 0.026905598118901253, + "grad_norm": 0.2511071562767029, + "learning_rate": 2.8838177686840577e-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": 0.013591368868947028, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.00630859131924808, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.03349132966250181, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.026905598118901253, + "weighted_rkd_loss": 0.016745664831250905 + }, + { + "epoch": 1.4700854700854702, + "flow_matching_loss": 0.028634503856301306, + "grad_norm": 0.3685453236103058, + "learning_rate": 2.8799591944442776e-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": 0.015366201288998127, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0075918316375464204, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.038324234448373316, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.028634503856301306, + "weighted_rkd_loss": 0.019162117224186658 + }, + { + "epoch": 1.4871794871794872, + "flow_matching_loss": 0.025696180015802383, + "grad_norm": 0.2295045554637909, + "learning_rate": 2.8760402563188848e-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": 0.014483218640089035, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.006810860987752676, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.035777297988533976, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.025696180015802383, + "weighted_rkd_loss": 0.017888648994266988 + }, + { + "epoch": 1.5042735042735043, + "flow_matching_loss": 0.02721292469650507, + "grad_norm": 0.2990417778491974, + "learning_rate": 2.8720611257349285e-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": 0.01320744464173913, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.006201161723583937, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.03261605091392994, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.02721292469650507, + "weighted_rkd_loss": 0.01630802545696497 + }, + { + "epoch": 1.5213675213675213, + "flow_matching_loss": 0.028353206627070904, + "grad_norm": 0.2800433337688446, + "learning_rate": 2.868021976752471e-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": 0.01263526426628232, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.006006072461605072, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.031276601180434224, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.028353206627070904, + "weighted_rkd_loss": 0.015638300590217112 + }, + { + "epoch": 1.5384615384615383, + "flow_matching_loss": 0.026863467507064344, + "grad_norm": 0.2563413977622986, + "learning_rate": 2.8639229860569738e-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": 0.01446392321959138, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.006428878894075752, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.03535672556608915, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.026863467507064344, + "weighted_rkd_loss": 0.017678362783044577 + }, + { + "epoch": 1.5555555555555556, + "flow_matching_loss": 0.025768650881946086, + "grad_norm": 0.29122796654701233, + "learning_rate": 2.8597643329515688e-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": 0.0123795872554183, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.005432371981441975, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.03019154630601406, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.025768650881946086, + "weighted_rkd_loss": 0.01509577315300703 + }, + { + "epoch": 1.5726495726495726, + "flow_matching_loss": 0.02874456774443388, + "grad_norm": 0.22805574536323547, + "learning_rate": 2.855546199349214e-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": 0.01520231394097209, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.006465561571530998, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.0368701895698905, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.02874456774443388, + "weighted_rkd_loss": 0.01843509478494525 + }, + { + "epoch": 1.5897435897435899, + "flow_matching_loss": 0.024948613718152046, + "grad_norm": 0.2613106667995453, + "learning_rate": 2.85126876976474e-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": 0.014754869975149632, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.00696990373544395, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.036479643732309344, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.024948613718152046, + "weighted_rkd_loss": 0.018239821866154672 + }, + { + "epoch": 1.606837606837607, + "flow_matching_loss": 0.025769425369799138, + "grad_norm": 0.2636370062828064, + "learning_rate": 2.8469322313067726e-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": 0.013760981895029546, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.006404791539534926, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.03392675518989563, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.025769425369799138, + "weighted_rkd_loss": 0.016963377594947815 + }, + { + "epoch": 1.623931623931624, + "flow_matching_loss": 0.02751282975077629, + "grad_norm": 0.23571224510669708, + "learning_rate": 2.8425367736695522e-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": 0.013957356754690408, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.006522616697475314, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.03443732969462872, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.02751282975077629, + "weighted_rkd_loss": 0.01721866484731436 + }, + { + "epoch": 1.641025641025641, + "flow_matching_loss": 0.024373013526201248, + "grad_norm": 0.2344050258398056, + "learning_rate": 2.838082589124636e-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": 0.012718899268656969, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.006056894036009908, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.03149469289928675, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.024373013526201248, + "weighted_rkd_loss": 0.015747346449643373 + }, + { + "epoch": 1.658119658119658, + "flow_matching_loss": 0.02341695725917816, + "grad_norm": 0.22509969770908356, + "learning_rate": 2.8335698725124843e-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": 0.014082430861890316, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.006729463767260313, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.03489432521164417, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.02341695725917816, + "weighted_rkd_loss": 0.017447162605822086 + }, + { + "epoch": 1.6752136752136753, + "flow_matching_loss": 0.022877768240869047, + "grad_norm": 0.2847081422805786, + "learning_rate": 2.8289988212339422e-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": 0.015481962030753494, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.006615176028572023, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.037579100392758844, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.022877768240869047, + "weighted_rkd_loss": 0.018789550196379422 + }, + { + "epoch": 1.6923076923076923, + "flow_matching_loss": 0.02152910204604268, + "grad_norm": 0.23003964126110077, + "learning_rate": 2.8243696352415997e-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": 0.014357429277151824, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.006650971900671721, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.03536583054810762, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.02152910204604268, + "weighted_rkd_loss": 0.01768291527405381 + }, + { + "epoch": 1.7094017094017095, + "flow_matching_loss": 0.027134731784462927, + "grad_norm": 0.2685432732105255, + "learning_rate": 2.8196825170310495e-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": 0.012607343588024377, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.005924688838422299, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.03113937582820654, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.027134731784462927, + "weighted_rkd_loss": 0.01556968791410327 + } + ], + "logging_steps": 10, + "max_steps": 5000, + "num_input_tokens_seen": 0, + "num_train_epochs": 9, + "save_steps": 5001, + "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": 64, + "trial_name": null, + "trial_params": null +} diff --git a/FewShot/39tasks/5shot/FromRKD-Timewarp_FullactionHead/seed2/default/train_5shot_seed2_fullactionhead_5k/checkpoint-3000/config.json b/FewShot/39tasks/5shot/FromRKD-Timewarp_FullactionHead/seed2/default/train_5shot_seed2_fullactionhead_5k/checkpoint-3000/config.json new file mode 100644 index 0000000000000000000000000000000000000000..4740d0455bef31c469ee5dab6d4a47d64df61665 --- /dev/null +++ b/FewShot/39tasks/5shot/FromRKD-Timewarp_FullactionHead/seed2/default/train_5shot_seed2_fullactionhead_5k/checkpoint-3000/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": 0.5, + "rkd_loss_weight_end": 0.0, + "rkd_loss_weight_schedule": null, + "rkd_loss_weight_start": 0.1, + "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/FewShot/39tasks/5shot/FromRKD-Timewarp_FullactionHead/seed2/default/train_5shot_seed2_fullactionhead_5k/checkpoint-3000/experiment_cfg/metadata.json b/FewShot/39tasks/5shot/FromRKD-Timewarp_FullactionHead/seed2/default/train_5shot_seed2_fullactionhead_5k/checkpoint-3000/experiment_cfg/metadata.json new file mode 100644 index 0000000000000000000000000000000000000000..23dcfa871a88e56280f945357aef0296bd071711 --- /dev/null +++ b/FewShot/39tasks/5shot/FromRKD-Timewarp_FullactionHead/seed2/default/train_5shot_seed2_fullactionhead_5k/checkpoint-3000/experiment_cfg/metadata.json @@ -0,0 +1,431 @@ +{ + "new_embodiment": { + "statistics": { + "state": { + "base_position": { + "max": [ + 7.778976917266846, + 0.33403685688972473, + 0.7223060727119446 + ], + "min": [ + -4.821934223175049, + -6.785110950469971, + 0.6994727253913879 + ], + "mean": [ + 2.577927505176127, + -2.1525618086425604, + 0.7003541659101092 + ], + "std": [ + 1.6720365948940972, + 1.3768414767482617, + 0.0009102921174316669 + ], + "q01": [ + -0.7605057497562686, + -5.518671947805291, + 0.6999764703640262 + ], + "q99": [ + 6.038879433964306, + -0.5168299981147954, + 0.7029182461269848 + ] + }, + "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, + 0.24782371371751144, + 0.5971093380149053 + ], + "std": [ + 0.0, + 0.0, + 0.6642559499983499, + 0.3751891785184018 + ], + "q01": [ + 0.0, + 0.0, + -1.0, + 4.194193576416142e-06 + ], + "q99": [ + 0.0, + 0.0, + 1.0, + 1.0 + ] + }, + "end_effector_position_relative": { + "max": [ + 0.8848381638526917, + 0.8569459915161133, + 1.053884744644165 + ], + "min": [ + -0.5828549861907959, + -0.8486236333847046, + -0.3446122705936432 + ], + "mean": [ + 0.25006971579930243, + -0.005716176032514759, + 0.4426027721228362 + ], + "std": [ + 0.15838384751435522, + 0.25141504139800513, + 0.20646816234620438 + ], + "q01": [ + -0.017003721077380903, + -0.47738320165276543, + 0.1917563409753314 + ], + "q99": [ + 0.5401029199137674, + 0.4706646609268599, + 0.6965586537529516 + ] + }, + "end_effector_rotation_relative": { + "max": [ + 0.9999949932098389, + 0.9997354745864868, + 0.9900747537612915, + 0.9735639095306396 + ], + "min": [ + -0.9999914169311523, + -0.9999884366989136, + -0.9975369572639465, + 6.754255821306288e-08 + ], + "mean": [ + -0.22531628862537245, + 0.045669797538445966, + -0.06098055586474368, + 0.12292005493282424 + ], + "std": [ + 0.8506341437448487, + 0.2880763358322842, + 0.30816573781739026, + 0.16366165154486362 + ], + "q01": [ + -0.9957510593732032, + -0.5584741016891311, + -0.5371647381159943, + 0.0013784815838340492 + ], + "q99": [ + 0.9899375209218226, + 0.6327878884078038, + 0.5071539024531994, + 0.43555975763914345 + ] + }, + "gripper_qpos": { + "max": [ + 0.05289580300450325, + 0.00937709677964449 + ], + "min": [ + -0.011669117026031017, + -0.06307756155729294 + ], + "mean": [ + 0.03196506153371015, + -0.03229791582989581 + ], + "std": [ + 0.01171640503534049, + 0.01140016755425199 + ], + "q01": [ + 0.006290779996034005, + -0.04059717954551848 + ], + "q99": [ + 0.04038165838477291, + -0.006363076424932449 + ] + } + }, + "action": { + "base_motion": { + "max": [ + 1.0, + 1.0, + 1.0, + 0.0 + ], + "min": [ + -1.0, + -1.0, + -1.0, + 0.0 + ], + "mean": [ + 0.006112990930475949, + 0.002504955933349057, + -0.001639409958610569, + 0.0 + ], + "std": [ + 0.11478077265783658, + 0.1002662368589144, + 0.0807258815139021, + 0.0 + ], + "q01": [ + -0.16969970921613756, + -0.19919537649977875, + -0.15012139866371593, + 0.0 + ], + "q99": [ + 0.2466625453958827, + 0.25040842772679267, + 0.12777701094549537, + 0.0 + ] + }, + "control_mode": { + "max": [ + 1.0 + ], + "min": [ + -1.0 + ], + "mean": [ + -0.9253067479599785 + ], + "std": [ + 0.3792397014797737 + ], + "q01": [ + -1.0 + ], + "q99": [ + -0.1974093068624371 + ] + }, + "end_effector_position": { + "max": [ + 1.0, + 1.0, + 1.0 + ], + "min": [ + -1.0, + -1.0, + -1.0 + ], + "mean": [ + -0.006754222051360652, + 0.0027854853184013867, + -0.08087572130330477 + ], + "std": [ + 0.45779905420789546, + 0.40578975219522756, + 0.4062694098132682 + ], + "q01": [ + -0.968295960290036, + -0.9240402337368848, + -0.8249359802202183 + ], + "q99": [ + 0.8046634915851651, + 0.9212243025654695, + 0.838433711437593 + ] + }, + "end_effector_rotation": { + "max": [ + 1.0, + 1.0, + 1.0 + ], + "min": [ + -1.0, + -1.0, + -1.0 + ], + "mean": [ + 0.008633749251857347, + -0.020406457524911718, + -0.0027639446362883006 + ], + "std": [ + 0.11393972872407104, + 0.13300186268009506, + 0.12192288302920105 + ], + "q01": [ + -0.2668579353776187, + -0.40299251206771697, + -0.34712957892172586 + ], + "q99": [ + 0.32748281264383927, + 0.3179028969292779, + 0.33459405510046736 + ] + }, + "gripper_close": { + "max": [ + 1.0 + ], + "min": [ + -1.0 + ], + "mean": [ + -0.3091635769453286 + ], + "std": [ + 0.9510122432876397 + ], + "q01": [ + -1.0 + ], + "q99": [ + 0.7643545615216344 + ] + } + } + }, + "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/FewShot/39tasks/5shot/FromRKD-Timewarp_FullactionHead/seed2/default/train_5shot_seed2_fullactionhead_5k/checkpoint-3000/model.safetensors.index.json b/FewShot/39tasks/5shot/FromRKD-Timewarp_FullactionHead/seed2/default/train_5shot_seed2_fullactionhead_5k/checkpoint-3000/model.safetensors.index.json new file mode 100644 index 0000000000000000000000000000000000000000..b20baed71f4d16a411223b70d45d934c6cbbfe97 --- /dev/null +++ b/FewShot/39tasks/5shot/FromRKD-Timewarp_FullactionHead/seed2/default/train_5shot_seed2_fullactionhead_5k/checkpoint-3000/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/FewShot/39tasks/5shot/FromRKD-Timewarp_FullactionHead/seed2/default/train_5shot_seed2_fullactionhead_5k/checkpoint-3000/trainer_state.json b/FewShot/39tasks/5shot/FromRKD-Timewarp_FullactionHead/seed2/default/train_5shot_seed2_fullactionhead_5k/checkpoint-3000/trainer_state.json new file mode 100644 index 0000000000000000000000000000000000000000..739a51cae79568e6fb35fc7485c65dfba7e67aeb --- /dev/null +++ b/FewShot/39tasks/5shot/FromRKD-Timewarp_FullactionHead/seed2/default/train_5shot_seed2_fullactionhead_5k/checkpoint-3000/trainer_state.json @@ -0,0 +1,13834 @@ +{ + "best_global_step": null, + "best_metric": null, + "best_model_checkpoint": null, + "epoch": 5.128205128205128, + "eval_steps": 500, + "global_step": 3000, + "is_hyper_param_search": false, + "is_local_process_zero": true, + "is_world_process_zero": true, + "log_history": [ + { + "epoch": 0.017094017094017096, + "flow_matching_loss": 0.1910901501774788, + "grad_norm": 1.1204875707626343, + "learning_rate": 1.08e-06, + "loss": 0.393, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.14386700689792634, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.11605296358466148, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.4037869781255722, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.1910901501774788, + "weighted_rkd_loss": 0.2018934890627861 + }, + { + "epoch": 0.03418803418803419, + "flow_matching_loss": 0.19379425793886185, + "grad_norm": 1.295363187789917, + "learning_rate": 2.28e-06, + "loss": 0.3834, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.1352515198290348, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.10873031765222549, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.3792333573102951, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.19379425793886185, + "weighted_rkd_loss": 0.18961667865514756 + }, + { + "epoch": 0.05128205128205128, + "flow_matching_loss": 0.17340936958789827, + "grad_norm": 0.9476425051689148, + "learning_rate": 3.48e-06, + "loss": 0.3566, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.1321360796689987, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.10214361175894737, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.36641576886177063, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.17340936958789827, + "weighted_rkd_loss": 0.18320788443088531 + }, + { + "epoch": 0.06837606837606838, + "flow_matching_loss": 0.16260478422045707, + "grad_norm": 0.8357043862342834, + "learning_rate": 4.68e-06, + "loss": 0.3417, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.12940022349357605, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.09945235326886177, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.3582527995109558, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.16260478422045707, + "weighted_rkd_loss": 0.1791263997554779 + }, + { + "epoch": 0.08547008547008547, + "flow_matching_loss": 0.13773404359817504, + "grad_norm": 0.7658578157424927, + "learning_rate": 5.8800000000000005e-06, + "loss": 0.3112, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.12533388957381247, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.09635578840970993, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.3470235675573349, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.13773404359817504, + "weighted_rkd_loss": 0.17351178377866744 + }, + { + "epoch": 0.10256410256410256, + "flow_matching_loss": 0.10675260052084923, + "grad_norm": 0.6615934371948242, + "learning_rate": 7.08e-06, + "loss": 0.2707, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.11958464086055756, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.08864015787839889, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.3278094381093979, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.10675260052084923, + "weighted_rkd_loss": 0.16390471905469894 + }, + { + "epoch": 0.11965811965811966, + "flow_matching_loss": 0.09638643190264702, + "grad_norm": 0.6668271422386169, + "learning_rate": 8.28e-06, + "loss": 0.2454, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.11045710295438767, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.07702444642782211, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.2979386538267136, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.09638643190264702, + "weighted_rkd_loss": 0.1489693269133568 + }, + { + "epoch": 0.13675213675213677, + "flow_matching_loss": 0.08751716017723084, + "grad_norm": 0.7303902506828308, + "learning_rate": 9.48e-06, + "loss": 0.2274, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.10436114817857742, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.07100482136011124, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.2797271177172661, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.08751716017723084, + "weighted_rkd_loss": 0.13986355885863305 + }, + { + "epoch": 0.15384615384615385, + "flow_matching_loss": 0.07772385030984878, + "grad_norm": 0.7202248573303223, + "learning_rate": 1.068e-05, + "loss": 0.2108, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.09959760904312134, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.06698438599705696, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.2661796033382416, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.07772385030984878, + "weighted_rkd_loss": 0.1330898016691208 + }, + { + "epoch": 0.17094017094017094, + "flow_matching_loss": 0.07342424914240837, + "grad_norm": 0.5669143199920654, + "learning_rate": 1.1880000000000001e-05, + "loss": 0.1941, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.09150870665907859, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.05835277736186981, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.24137018769979476, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.07342424914240837, + "weighted_rkd_loss": 0.12068509384989738 + }, + { + "epoch": 0.18803418803418803, + "flow_matching_loss": 0.07115557119250297, + "grad_norm": 0.5177780985832214, + "learning_rate": 1.308e-05, + "loss": 0.1825, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.08484098091721534, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.05309118740260601, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.22277315109968185, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.07115557119250297, + "weighted_rkd_loss": 0.11138657554984092 + }, + { + "epoch": 0.20512820512820512, + "flow_matching_loss": 0.06366660706698894, + "grad_norm": 0.6365015506744385, + "learning_rate": 1.428e-05, + "loss": 0.1748, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.08515534773468972, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0519100297242403, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.2222207233309746, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.06366660706698894, + "weighted_rkd_loss": 0.1111103616654873 + }, + { + "epoch": 0.2222222222222222, + "flow_matching_loss": 0.06294491589069366, + "grad_norm": 0.5627866983413696, + "learning_rate": 1.548e-05, + "loss": 0.165, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.07825233079493046, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.04756654389202595, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.20407120436429976, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.06294491589069366, + "weighted_rkd_loss": 0.10203560218214988 + }, + { + "epoch": 0.23931623931623933, + "flow_matching_loss": 0.061772428825497626, + "grad_norm": 0.6260324716567993, + "learning_rate": 1.6680000000000003e-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": 0.0717402745038271, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.04264889135956764, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.18612944036722184, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.061772428825497626, + "weighted_rkd_loss": 0.09306472018361092 + }, + { + "epoch": 0.2564102564102564, + "flow_matching_loss": 0.061649881303310394, + "grad_norm": 0.6016267538070679, + "learning_rate": 1.7879999999999998e-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": 0.06565237119793892, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.03857212290167809, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.16987686604261398, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.061649881303310394, + "weighted_rkd_loss": 0.08493843302130699 + }, + { + "epoch": 0.27350427350427353, + "flow_matching_loss": 0.06200782023370266, + "grad_norm": 0.5819275379180908, + "learning_rate": 1.908e-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": 0.06578411012887955, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.03858097940683365, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.17014919966459274, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.06200782023370266, + "weighted_rkd_loss": 0.08507459983229637 + }, + { + "epoch": 0.2905982905982906, + "flow_matching_loss": 0.05749040208756924, + "grad_norm": 0.537067711353302, + "learning_rate": 2.0280000000000002e-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": 0.06401401236653328, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.036967079155147076, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.16499510556459426, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.05749040208756924, + "weighted_rkd_loss": 0.08249755278229713 + }, + { + "epoch": 0.3076923076923077, + "flow_matching_loss": 0.05535002276301384, + "grad_norm": 0.4923454225063324, + "learning_rate": 2.148e-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": 0.05838514976203442, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.033644383773207664, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.1504146859049797, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.05535002276301384, + "weighted_rkd_loss": 0.07520734295248985 + }, + { + "epoch": 0.3247863247863248, + "flow_matching_loss": 0.05286722965538502, + "grad_norm": 0.48796457052230835, + "learning_rate": 2.268e-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": 0.05483606085181236, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.031941465102136135, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.1416135847568512, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.05286722965538502, + "weighted_rkd_loss": 0.0708067923784256 + }, + { + "epoch": 0.3418803418803419, + "flow_matching_loss": 0.05254113264381886, + "grad_norm": 0.4936208128929138, + "learning_rate": 2.3880000000000002e-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": 0.05183771774172783, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.029638106748461724, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.13331354185938835, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.05254113264381886, + "weighted_rkd_loss": 0.06665677092969417 + }, + { + "epoch": 0.358974358974359, + "flow_matching_loss": 0.04968957602977753, + "grad_norm": 0.5536408424377441, + "learning_rate": 2.508e-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": 0.047118809819221494, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.02699419390410185, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.12123181372880935, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.04968957602977753, + "weighted_rkd_loss": 0.06061590686440468 + }, + { + "epoch": 0.37606837606837606, + "flow_matching_loss": 0.04866243675351143, + "grad_norm": 0.48103460669517517, + "learning_rate": 2.628e-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": 0.046376798301935196, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.02550226878374815, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.11825586706399918, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.04866243675351143, + "weighted_rkd_loss": 0.05912793353199959 + }, + { + "epoch": 0.39316239316239315, + "flow_matching_loss": 0.04957115277647972, + "grad_norm": 0.5273743271827698, + "learning_rate": 2.748e-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": 0.04591682404279709, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.02485441230237484, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.11668805927038192, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.04957115277647972, + "weighted_rkd_loss": 0.05834402963519096 + }, + { + "epoch": 0.41025641025641024, + "flow_matching_loss": 0.04972734041512013, + "grad_norm": 0.48441460728645325, + "learning_rate": 2.868e-05, + "loss": 0.105, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.04358960241079331, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.023280760645866393, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.11045996621251106, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.04972734041512013, + "weighted_rkd_loss": 0.05522998310625553 + }, + { + "epoch": 0.42735042735042733, + "flow_matching_loss": 0.046684912964701654, + "grad_norm": 0.45080485939979553, + "learning_rate": 2.9880000000000002e-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": 0.04428564272820949, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.02520337775349617, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.1137746624648571, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.046684912964701654, + "weighted_rkd_loss": 0.05688733123242855 + }, + { + "epoch": 0.4444444444444444, + "flow_matching_loss": 0.04779224582016468, + "grad_norm": 0.43459874391555786, + "learning_rate": 2.9999734259635676e-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": 0.036030808463692665, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.019671429321169852, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.0917330451309681, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.04779224582016468, + "weighted_rkd_loss": 0.04586652256548405 + }, + { + "epoch": 0.46153846153846156, + "flow_matching_loss": 0.04915213659405708, + "grad_norm": 0.4534752368927002, + "learning_rate": 2.9998815663057245e-05, + "loss": 0.0983, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.039135459996759894, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.01992818871513009, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.09819910898804665, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.04915213659405708, + "weighted_rkd_loss": 0.049099554494023326 + }, + { + "epoch": 0.47863247863247865, + "flow_matching_loss": 0.044950670748949054, + "grad_norm": 0.4854293167591095, + "learning_rate": 2.999724096969228e-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": 0.03963045179843903, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.02123024109750986, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.10049114525318145, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.044950670748949054, + "weighted_rkd_loss": 0.050245572626590726 + }, + { + "epoch": 0.49572649572649574, + "flow_matching_loss": 0.045206136628985406, + "grad_norm": 0.4032607972621918, + "learning_rate": 2.9995010248422984e-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": 0.03566844779998064, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.018644866067916155, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.08998176082968712, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.045206136628985406, + "weighted_rkd_loss": 0.04499088041484356 + }, + { + "epoch": 0.5128205128205128, + "flow_matching_loss": 0.042293429002165796, + "grad_norm": 0.4287409782409668, + "learning_rate": 2.999212359682832e-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": 0.03817300200462341, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.020988442189991473, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.09733444601297378, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.042293429002165796, + "weighted_rkd_loss": 0.04866722300648689 + }, + { + "epoch": 0.5299145299145299, + "flow_matching_loss": 0.04139162935316563, + "grad_norm": 0.45648202300071716, + "learning_rate": 2.9988581141179787e-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": 0.03135862443596125, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.016660952381789683, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.07937820106744767, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.04139162935316563, + "weighted_rkd_loss": 0.039689100533723834 + }, + { + "epoch": 0.5470085470085471, + "flow_matching_loss": 0.041223404183983806, + "grad_norm": 0.4205642342567444, + "learning_rate": 2.998438303643587e-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": 0.034546280093491075, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.018318970408290624, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.08741152957081795, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.041223404183983806, + "weighted_rkd_loss": 0.043705764785408975 + }, + { + "epoch": 0.5641025641025641, + "flow_matching_loss": 0.04060795344412327, + "grad_norm": 0.402844101190567, + "learning_rate": 2.997952946623527e-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": 0.0335417652502656, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.018018722627311946, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.08510225303471089, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.04060795344412327, + "weighted_rkd_loss": 0.042551126517355445 + }, + { + "epoch": 0.5811965811965812, + "flow_matching_loss": 0.04365103021264076, + "grad_norm": 0.37454837560653687, + "learning_rate": 2.9974020642888875e-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": 0.03194001261144876, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.016204175911843775, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.08008420169353485, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.04365103021264076, + "weighted_rkd_loss": 0.040042100846767424 + }, + { + "epoch": 0.5982905982905983, + "flow_matching_loss": 0.039248546957969664, + "grad_norm": 0.38750800490379333, + "learning_rate": 2.9967856807370458e-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": 0.03014408703893423, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.015359362401068211, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.07564753592014313, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.039248546957969664, + "weighted_rkd_loss": 0.037823767960071565 + }, + { + "epoch": 0.6153846153846154, + "flow_matching_loss": 0.04174409061670303, + "grad_norm": 0.4128570556640625, + "learning_rate": 2.996103822930615e-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": 0.030098697915673257, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.015234342124313117, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.07543173767626285, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.04174409061670303, + "weighted_rkd_loss": 0.037715868838131425 + }, + { + "epoch": 0.6324786324786325, + "flow_matching_loss": 0.038203606754541396, + "grad_norm": 0.3553514778614044, + "learning_rate": 2.9953565206962653e-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": 0.03010321594774723, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.015880199056118725, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.07608662992715835, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.038203606754541396, + "weighted_rkd_loss": 0.038043314963579176 + }, + { + "epoch": 0.6495726495726496, + "flow_matching_loss": 0.038802698254585266, + "grad_norm": 0.393643319606781, + "learning_rate": 2.9945438067234172e-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": 0.02764769084751606, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.013171730842441321, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.06846711188554763, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.038802698254585266, + "weighted_rkd_loss": 0.034233555942773816 + }, + { + "epoch": 0.6666666666666666, + "flow_matching_loss": 0.03741585947573185, + "grad_norm": 0.38524219393730164, + "learning_rate": 2.9936657165628124e-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": 0.0301553001627326, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.015053990110754967, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.07536459155380726, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.03741585947573185, + "weighted_rkd_loss": 0.03768229577690363 + }, + { + "epoch": 0.6837606837606838, + "flow_matching_loss": 0.037859869003295896, + "grad_norm": 0.3669644594192505, + "learning_rate": 2.992722288624959e-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": 0.02703232318162918, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.014790859073400497, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.06885550543665886, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.037859869003295896, + "weighted_rkd_loss": 0.03442775271832943 + }, + { + "epoch": 0.7008547008547008, + "flow_matching_loss": 0.03818879183381796, + "grad_norm": 0.30247363448143005, + "learning_rate": 2.9917135641784515e-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": 0.026851335540413857, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.013741841074079275, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.06744451224803924, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.03818879183381796, + "weighted_rkd_loss": 0.03372225612401962 + }, + { + "epoch": 0.717948717948718, + "flow_matching_loss": 0.0374411903321743, + "grad_norm": 0.4039696753025055, + "learning_rate": 2.9906395873481655e-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": 0.024242410250008107, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.012942939531058073, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.061427760496735574, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.0374411903321743, + "weighted_rkd_loss": 0.030713880248367787 + }, + { + "epoch": 0.7350427350427351, + "flow_matching_loss": 0.0361711286008358, + "grad_norm": 0.34961435198783875, + "learning_rate": 2.989500405113326e-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": 0.024853001721203326, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.011823360808193684, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.061529364064335824, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.0361711286008358, + "weighted_rkd_loss": 0.030764682032167912 + }, + { + "epoch": 0.7521367521367521, + "flow_matching_loss": 0.03577544335275888, + "grad_norm": 0.3615719676017761, + "learning_rate": 2.9882960673054547e-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": 0.02502665389329195, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.01260580299422145, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.06265911161899566, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.03577544335275888, + "weighted_rkd_loss": 0.03132955580949783 + }, + { + "epoch": 0.7692307692307693, + "flow_matching_loss": 0.03441168405115604, + "grad_norm": 0.3467264175415039, + "learning_rate": 2.9870266266061886e-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": 0.02530351039022207, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.011560741625726224, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.062167763337492944, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.03441168405115604, + "weighted_rkd_loss": 0.031083881668746472 + }, + { + "epoch": 0.7863247863247863, + "flow_matching_loss": 0.03624687343835831, + "grad_norm": 0.3562852442264557, + "learning_rate": 2.9856921385449766e-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": 0.025862449035048484, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.012488656770437956, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.06421355418860912, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.03624687343835831, + "weighted_rkd_loss": 0.03210677709430456 + }, + { + "epoch": 0.8034188034188035, + "flow_matching_loss": 0.03741344213485718, + "grad_norm": 0.3594230115413666, + "learning_rate": 2.9842926614966502e-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": 0.021881727501749992, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.01110158571973443, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.054865040257573126, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.03741344213485718, + "weighted_rkd_loss": 0.027432520128786563 + }, + { + "epoch": 0.8205128205128205, + "flow_matching_loss": 0.029932724311947823, + "grad_norm": 0.3690202832221985, + "learning_rate": 2.982828256678868e-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": 0.02407331857830286, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.012713254615664481, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.060859891772270205, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.029932724311947823, + "weighted_rkd_loss": 0.030429945886135103 + }, + { + "epoch": 0.8376068376068376, + "flow_matching_loss": 0.03617270402610302, + "grad_norm": 0.38251274824142456, + "learning_rate": 2.9812989881494412e-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": 0.025536379404366018, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.012556855008006096, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.06362961418926716, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.03617270402610302, + "weighted_rkd_loss": 0.03181480709463358 + }, + { + "epoch": 0.8547008547008547, + "flow_matching_loss": 0.03683352377265692, + "grad_norm": 0.3576796352863312, + "learning_rate": 2.9797049228035296e-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": 0.02277378421276808, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.011028627213090659, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.056576195359230044, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.03683352377265692, + "weighted_rkd_loss": 0.028288097679615022 + }, + { + "epoch": 0.8717948717948718, + "flow_matching_loss": 0.034184624627232554, + "grad_norm": 0.4046909511089325, + "learning_rate": 2.978046130370715e-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": 0.02267690245062113, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.011717867571860552, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.05707167312502861, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.034184624627232554, + "weighted_rkd_loss": 0.028535836562514304 + }, + { + "epoch": 0.8888888888888888, + "flow_matching_loss": 0.03213700447231531, + "grad_norm": 0.3536110818386078, + "learning_rate": 2.9763226834119535e-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": 0.02117164060473442, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.01059307949617505, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.052936359867453577, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.03213700447231531, + "weighted_rkd_loss": 0.026468179933726788 + }, + { + "epoch": 0.905982905982906, + "flow_matching_loss": 0.03486904371529818, + "grad_norm": 0.3733022212982178, + "learning_rate": 2.9745346573163963e-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": 0.02185452114790678, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.01133495531976223, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.05504399687051773, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.03486904371529818, + "weighted_rkd_loss": 0.027521998435258866 + }, + { + "epoch": 0.9230769230769231, + "flow_matching_loss": 0.031505046598613265, + "grad_norm": 0.35847941040992737, + "learning_rate": 2.9726821302980987e-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": 0.020499915070831775, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.01087835356593132, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.05187818445265293, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.031505046598613265, + "weighted_rkd_loss": 0.025939092226326465 + }, + { + "epoch": 0.9401709401709402, + "flow_matching_loss": 0.030537511594593524, + "grad_norm": 0.3062575161457062, + "learning_rate": 2.9707651833925935e-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": 0.019180506095290185, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.009259850112721324, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.04762086160480976, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.030537511594593524, + "weighted_rkd_loss": 0.02381043080240488 + }, + { + "epoch": 0.9572649572649573, + "flow_matching_loss": 0.033962773904204366, + "grad_norm": 0.33466994762420654, + "learning_rate": 2.968783900453348e-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": 0.019049630872905253, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.009260836755856872, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.04736009836196899, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.033962773904204366, + "weighted_rkd_loss": 0.023680049180984496 + }, + { + "epoch": 0.9743589743589743, + "flow_matching_loss": 0.03441021610051394, + "grad_norm": 0.3329583704471588, + "learning_rate": 2.9667383681480968e-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": 0.02119819391518831, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.00991179826669395, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.052308186143636706, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.03441021610051394, + "weighted_rkd_loss": 0.026154093071818353 + }, + { + "epoch": 0.9914529914529915, + "flow_matching_loss": 0.032573466189205647, + "grad_norm": 0.3133791387081146, + "learning_rate": 2.9646286759550493e-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": 0.019760795310139655, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.00934667899273336, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.048868270218372346, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.032573466189205647, + "weighted_rkd_loss": 0.024434135109186173 + }, + { + "epoch": 1.0085470085470085, + "flow_matching_loss": 0.030896119587123395, + "grad_norm": 0.2655918002128601, + "learning_rate": 2.9624549161589764e-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": 0.018024975061416627, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.008287688577547669, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.04433763958513737, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.030896119587123395, + "weighted_rkd_loss": 0.022168819792568684 + }, + { + "epoch": 1.0256410256410255, + "flow_matching_loss": 0.02869145143777132, + "grad_norm": 0.2975919544696808, + "learning_rate": 2.9602171838471745e-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": 0.01704270886257291, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.008086427114903927, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.042171845026314256, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.02869145143777132, + "weighted_rkd_loss": 0.021085922513157128 + }, + { + "epoch": 1.0427350427350428, + "flow_matching_loss": 0.028883341886103153, + "grad_norm": 0.25782883167266846, + "learning_rate": 2.957915576905303e-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": 0.020831129886209963, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.01092052492313087, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.052582784369587895, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.028883341886103153, + "weighted_rkd_loss": 0.026291392184793948 + }, + { + "epoch": 1.0598290598290598, + "flow_matching_loss": 0.02891363874077797, + "grad_norm": 0.3048412501811981, + "learning_rate": 2.9555501960131058e-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": 0.018006673827767373, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.00887504410929978, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.044888391718268396, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.02891363874077797, + "weighted_rkd_loss": 0.022444195859134198 + }, + { + "epoch": 1.0769230769230769, + "flow_matching_loss": 0.029369746521115302, + "grad_norm": 0.3183818757534027, + "learning_rate": 2.9531211446400068e-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": 0.016618114057928325, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.007927177706733346, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.041163405403494835, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.029369746521115302, + "weighted_rkd_loss": 0.020581702701747417 + }, + { + "epoch": 1.0940170940170941, + "flow_matching_loss": 0.028417748771607874, + "grad_norm": 0.2978658378124237, + "learning_rate": 2.9506285290405813e-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": 0.01775301033630967, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.007838332653045654, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.04334435388445854, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.028417748771607874, + "weighted_rkd_loss": 0.02167217694222927 + }, + { + "epoch": 1.1111111111111112, + "flow_matching_loss": 0.030568437837064266, + "grad_norm": 0.27075809240341187, + "learning_rate": 2.948072458249911e-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": 0.01854327218607068, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.009335583634674549, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.04642212763428688, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.030568437837064266, + "weighted_rkd_loss": 0.02321106381714344 + }, + { + "epoch": 1.1282051282051282, + "flow_matching_loss": 0.03013926316052675, + "grad_norm": 0.31489095091819763, + "learning_rate": 2.9454530440788125e-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": 0.018000216316431762, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.008591329725459219, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.044591762498021124, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.03013926316052675, + "weighted_rkd_loss": 0.022295881249010562 + }, + { + "epoch": 1.1452991452991452, + "flow_matching_loss": 0.03240344282239675, + "grad_norm": 0.3196583092212677, + "learning_rate": 2.9427704011089474e-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": 0.01813620710745454, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.008372926386073232, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.044645340740680696, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.03240344282239675, + "weighted_rkd_loss": 0.022322670370340348 + }, + { + "epoch": 1.1623931623931625, + "flow_matching_loss": 0.02988831289112568, + "grad_norm": 0.33447059988975525, + "learning_rate": 2.94002464668781e-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": 0.017554160580039023, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.008155462937429548, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.04326378367841244, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.02988831289112568, + "weighted_rkd_loss": 0.02163189183920622 + }, + { + "epoch": 1.1794871794871795, + "flow_matching_loss": 0.0297787019982934, + "grad_norm": 0.25146859884262085, + "learning_rate": 2.9372159009235936e-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": 0.016257153637707234, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.007763447891920805, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.040277755446732046, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.0297787019982934, + "weighted_rkd_loss": 0.020138877723366023 + }, + { + "epoch": 1.1965811965811965, + "flow_matching_loss": 0.033523568883538246, + "grad_norm": 0.2588149905204773, + "learning_rate": 2.9343442866799365e-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": 0.016528852004557847, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.007534453133121133, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.0405921570956707, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.033523568883538246, + "weighted_rkd_loss": 0.02029607854783535 + }, + { + "epoch": 1.2136752136752136, + "flow_matching_loss": 0.029541657492518424, + "grad_norm": 0.2806798219680786, + "learning_rate": 2.9314099295705495e-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": 0.017626770399510862, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.008789777662605048, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.04404331874102354, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.029541657492518424, + "weighted_rkd_loss": 0.02202165937051177 + }, + { + "epoch": 1.2307692307692308, + "flow_matching_loss": 0.02779527623206377, + "grad_norm": 0.23124118149280548, + "learning_rate": 2.928412957953718e-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": 0.017632605507969856, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.008616828918457031, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.043882039934396745, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.02779527623206377, + "weighted_rkd_loss": 0.021941019967198373 + }, + { + "epoch": 1.2478632478632479, + "flow_matching_loss": 0.028330446407198905, + "grad_norm": 0.24973063170909882, + "learning_rate": 2.925353502926689e-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": 0.015281323064118623, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.007210866408422589, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.037773511931300165, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.028330446407198905, + "weighted_rkd_loss": 0.018886755965650082 + }, + { + "epoch": 1.264957264957265, + "flow_matching_loss": 0.026782910339534283, + "grad_norm": 0.29868558049201965, + "learning_rate": 2.922231698319937e-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": 0.016518175043165682, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.008065681951120496, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.04110203199088573, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.026782910339534283, + "weighted_rkd_loss": 0.020551015995442866 + }, + { + "epoch": 1.282051282051282, + "flow_matching_loss": 0.026428927294909955, + "grad_norm": 0.2616300582885742, + "learning_rate": 2.9190476806913088e-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": 0.015521802846342326, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.007240159553475678, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.03828376531600952, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.026428927294909955, + "weighted_rkd_loss": 0.01914188265800476 + }, + { + "epoch": 1.2991452991452992, + "flow_matching_loss": 0.02718909680843353, + "grad_norm": 0.2325822114944458, + "learning_rate": 2.9158015893200504e-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": 0.016340584587305786, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.006903730751946569, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.03958490006625652, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.02718909680843353, + "weighted_rkd_loss": 0.01979245003312826 + }, + { + "epoch": 1.3162393162393162, + "flow_matching_loss": 0.028375131264328956, + "grad_norm": 0.32508528232574463, + "learning_rate": 2.9124935662007137e-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": 0.015554658975452185, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.007059313822537661, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.03816863112151623, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.028375131264328956, + "weighted_rkd_loss": 0.019084315560758115 + }, + { + "epoch": 1.3333333333333333, + "flow_matching_loss": 0.027656448632478715, + "grad_norm": 0.27773064374923706, + "learning_rate": 2.9091237560369467e-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": 0.01565126357600093, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.007432594196870923, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.03873512130230665, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.027656448632478715, + "weighted_rkd_loss": 0.019367560651153325 + }, + { + "epoch": 1.3504273504273505, + "flow_matching_loss": 0.02753492761403322, + "grad_norm": 0.2521279454231262, + "learning_rate": 2.905692306235163e-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": 0.014433691278100014, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0070426968391984705, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.03591007981449366, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.02753492761403322, + "weighted_rkd_loss": 0.01795503990724683 + }, + { + "epoch": 1.3675213675213675, + "flow_matching_loss": 0.027198122441768648, + "grad_norm": 0.2660417854785919, + "learning_rate": 2.9021993668980925e-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": 0.015085380058735609, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.007494675228372216, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.03766543548554182, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.027198122441768648, + "weighted_rkd_loss": 0.01883271774277091 + }, + { + "epoch": 1.3846153846153846, + "flow_matching_loss": 0.031527518108487126, + "grad_norm": 0.3124120831489563, + "learning_rate": 2.8986450908182185e-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": 0.015878252685070038, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.007188629452139139, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.038945135287940504, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.031527518108487126, + "weighted_rkd_loss": 0.019472567643970252 + }, + { + "epoch": 1.4017094017094016, + "flow_matching_loss": 0.028070161677896978, + "grad_norm": 0.23588840663433075, + "learning_rate": 2.8950296334710902e-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": 0.016684133745729923, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.007126115215942264, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.040494382940232755, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.028070161677896978, + "weighted_rkd_loss": 0.020247191470116378 + }, + { + "epoch": 1.4188034188034189, + "flow_matching_loss": 0.02908680960536003, + "grad_norm": 0.24586918950080872, + "learning_rate": 2.8913531530085248e-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": 0.014786572381854057, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.007374308165162801, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.036947453022003175, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.02908680960536003, + "weighted_rkd_loss": 0.018473726511001588 + }, + { + "epoch": 1.435897435897436, + "flow_matching_loss": 0.026878987438976765, + "grad_norm": 0.22994333505630493, + "learning_rate": 2.887615810251687e-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": 0.01679125111550093, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.007170996256172657, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.040753497555851934, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.026878987438976765, + "weighted_rkd_loss": 0.020376748777925967 + }, + { + "epoch": 1.452991452991453, + "flow_matching_loss": 0.026905598118901253, + "grad_norm": 0.2511071562767029, + "learning_rate": 2.8838177686840577e-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": 0.013591368868947028, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.00630859131924808, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.03349132966250181, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.026905598118901253, + "weighted_rkd_loss": 0.016745664831250905 + }, + { + "epoch": 1.4700854700854702, + "flow_matching_loss": 0.028634503856301306, + "grad_norm": 0.3685453236103058, + "learning_rate": 2.8799591944442776e-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": 0.015366201288998127, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0075918316375464204, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.038324234448373316, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.028634503856301306, + "weighted_rkd_loss": 0.019162117224186658 + }, + { + "epoch": 1.4871794871794872, + "flow_matching_loss": 0.025696180015802383, + "grad_norm": 0.2295045554637909, + "learning_rate": 2.8760402563188848e-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": 0.014483218640089035, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.006810860987752676, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.035777297988533976, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.025696180015802383, + "weighted_rkd_loss": 0.017888648994266988 + }, + { + "epoch": 1.5042735042735043, + "flow_matching_loss": 0.02721292469650507, + "grad_norm": 0.2990417778491974, + "learning_rate": 2.8720611257349285e-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": 0.01320744464173913, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.006201161723583937, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.03261605091392994, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.02721292469650507, + "weighted_rkd_loss": 0.01630802545696497 + }, + { + "epoch": 1.5213675213675213, + "flow_matching_loss": 0.028353206627070904, + "grad_norm": 0.2800433337688446, + "learning_rate": 2.868021976752471e-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": 0.01263526426628232, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.006006072461605072, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.031276601180434224, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.028353206627070904, + "weighted_rkd_loss": 0.015638300590217112 + }, + { + "epoch": 1.5384615384615383, + "flow_matching_loss": 0.026863467507064344, + "grad_norm": 0.2563413977622986, + "learning_rate": 2.8639229860569738e-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": 0.01446392321959138, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.006428878894075752, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.03535672556608915, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.026863467507064344, + "weighted_rkd_loss": 0.017678362783044577 + }, + { + "epoch": 1.5555555555555556, + "flow_matching_loss": 0.025768650881946086, + "grad_norm": 0.29122796654701233, + "learning_rate": 2.8597643329515688e-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": 0.0123795872554183, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.005432371981441975, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.03019154630601406, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.025768650881946086, + "weighted_rkd_loss": 0.01509577315300703 + }, + { + "epoch": 1.5726495726495726, + "flow_matching_loss": 0.02874456774443388, + "grad_norm": 0.22805574536323547, + "learning_rate": 2.855546199349214e-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": 0.01520231394097209, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.006465561571530998, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.0368701895698905, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.02874456774443388, + "weighted_rkd_loss": 0.01843509478494525 + }, + { + "epoch": 1.5897435897435899, + "flow_matching_loss": 0.024948613718152046, + "grad_norm": 0.2613106667995453, + "learning_rate": 2.85126876976474e-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": 0.014754869975149632, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.00696990373544395, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.036479643732309344, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.024948613718152046, + "weighted_rkd_loss": 0.018239821866154672 + }, + { + "epoch": 1.606837606837607, + "flow_matching_loss": 0.025769425369799138, + "grad_norm": 0.2636370062828064, + "learning_rate": 2.8469322313067726e-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": 0.013760981895029546, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.006404791539534926, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.03392675518989563, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.025769425369799138, + "weighted_rkd_loss": 0.016963377594947815 + }, + { + "epoch": 1.623931623931624, + "flow_matching_loss": 0.02751282975077629, + "grad_norm": 0.23571224510669708, + "learning_rate": 2.8425367736695522e-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": 0.013957356754690408, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.006522616697475314, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.03443732969462872, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.02751282975077629, + "weighted_rkd_loss": 0.01721866484731436 + }, + { + "epoch": 1.641025641025641, + "flow_matching_loss": 0.024373013526201248, + "grad_norm": 0.2344050258398056, + "learning_rate": 2.838082589124636e-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": 0.012718899268656969, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.006056894036009908, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.03149469289928675, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.024373013526201248, + "weighted_rkd_loss": 0.015747346449643373 + }, + { + "epoch": 1.658119658119658, + "flow_matching_loss": 0.02341695725917816, + "grad_norm": 0.22509969770908356, + "learning_rate": 2.8335698725124843e-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": 0.014082430861890316, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.006729463767260313, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.03489432521164417, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.02341695725917816, + "weighted_rkd_loss": 0.017447162605822086 + }, + { + "epoch": 1.6752136752136753, + "flow_matching_loss": 0.022877768240869047, + "grad_norm": 0.2847081422805786, + "learning_rate": 2.8289988212339422e-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": 0.015481962030753494, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.006615176028572023, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.037579100392758844, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.022877768240869047, + "weighted_rkd_loss": 0.018789550196379422 + }, + { + "epoch": 1.6923076923076923, + "flow_matching_loss": 0.02152910204604268, + "grad_norm": 0.23003964126110077, + "learning_rate": 2.8243696352415997e-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": 0.014357429277151824, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.006650971900671721, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.03536583054810762, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.02152910204604268, + "weighted_rkd_loss": 0.01768291527405381 + }, + { + "epoch": 1.7094017094017095, + "flow_matching_loss": 0.027134731784462927, + "grad_norm": 0.2685432732105255, + "learning_rate": 2.8196825170310495e-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": 0.012607343588024377, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.005924688838422299, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.03113937582820654, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.027134731784462927, + "weighted_rkd_loss": 0.01556968791410327 + }, + { + "epoch": 1.7264957264957266, + "flow_matching_loss": 0.024589456059038638, + "grad_norm": 0.3030104637145996, + "learning_rate": 2.8149376716320256e-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": 0.013603074103593826, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0063509512692689896, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.03355709929019213, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.024589456059038638, + "weighted_rkd_loss": 0.016778549645096064 + }, + { + "epoch": 1.7435897435897436, + "flow_matching_loss": 0.023589758202433586, + "grad_norm": 0.30140846967697144, + "learning_rate": 2.8101353065994373e-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": 0.013760259561240672, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.005739910667762161, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.03326043039560318, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.023589758202433586, + "weighted_rkd_loss": 0.01663021519780159 + }, + { + "epoch": 1.7606837606837606, + "flow_matching_loss": 0.023107477091252805, + "grad_norm": 0.32013994455337524, + "learning_rate": 2.8052756320042885e-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": 0.015008250717073679, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.007236170023679734, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.03725267145782709, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.023107477091252805, + "weighted_rkd_loss": 0.018626335728913545 + }, + { + "epoch": 1.7777777777777777, + "flow_matching_loss": 0.023336588591337203, + "grad_norm": 0.2473699301481247, + "learning_rate": 2.8003588604244893e-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": 0.013521139323711396, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.006724996492266655, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.03376727476716042, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.023336588591337203, + "weighted_rkd_loss": 0.01688363738358021 + }, + { + "epoch": 1.7948717948717947, + "flow_matching_loss": 0.02315734177827835, + "grad_norm": 0.26240313053131104, + "learning_rate": 2.7953852069355565e-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": 0.013256608508527279, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.005763246980495751, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.0322764640673995, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.02315734177827835, + "weighted_rkd_loss": 0.01613823203369975 + }, + { + "epoch": 1.811965811965812, + "flow_matching_loss": 0.024713453091681003, + "grad_norm": 0.2214726358652115, + "learning_rate": 2.790354889101206e-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": 0.010995917860418558, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.004965055221691728, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.026956891268491746, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.024713453091681003, + "weighted_rkd_loss": 0.013478445634245873 + }, + { + "epoch": 1.8290598290598292, + "flow_matching_loss": 0.02438421081751585, + "grad_norm": 0.2749342918395996, + "learning_rate": 2.7852681269638366e-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": 0.011696519609540702, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.005299521028064191, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.028692560084164143, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.02438421081751585, + "weighted_rkd_loss": 0.014346280042082072 + }, + { + "epoch": 1.8461538461538463, + "flow_matching_loss": 0.021150885801762342, + "grad_norm": 0.3021300137042999, + "learning_rate": 2.780125143034902e-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": 0.011429396085441112, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.005000467319041491, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.027859259769320487, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.021150885801762342, + "weighted_rkd_loss": 0.013929629884660243 + }, + { + "epoch": 1.8632478632478633, + "flow_matching_loss": 0.025476083718240262, + "grad_norm": 0.22854529321193695, + "learning_rate": 2.7749261622851812e-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": 0.011397741734981537, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.005167530756443739, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.027963014133274556, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.025476083718240262, + "weighted_rkd_loss": 0.013981507066637278 + }, + { + "epoch": 1.8803418803418803, + "flow_matching_loss": 0.025076989829540253, + "grad_norm": 0.2469537854194641, + "learning_rate": 2.7696714121349335e-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": 0.012464882060885429, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.005298179853707552, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.030227943509817123, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.025076989829540253, + "weighted_rkd_loss": 0.015113971754908561 + }, + { + "epoch": 1.8974358974358974, + "flow_matching_loss": 0.025070505402982234, + "grad_norm": 0.271639883518219, + "learning_rate": 2.764361122443954e-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": 0.013381752371788024, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.00690468696411699, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.03366819154471159, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.025070505402982234, + "weighted_rkd_loss": 0.016834095772355794 + }, + { + "epoch": 1.9145299145299144, + "flow_matching_loss": 0.021389586478471757, + "grad_norm": 0.2531188130378723, + "learning_rate": 2.758995525501517e-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": 0.012239555548876523, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.005215863836929202, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.029694975167512894, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.021389586478471757, + "weighted_rkd_loss": 0.014847487583756447 + }, + { + "epoch": 1.9316239316239316, + "flow_matching_loss": 0.025069457851350306, + "grad_norm": 0.2373230755329132, + "learning_rate": 2.753574856016215e-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": 0.010860348027199507, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.005091451085172593, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.0268121475353837, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.025069457851350306, + "weighted_rkd_loss": 0.01340607376769185 + }, + { + "epoch": 1.9487179487179487, + "flow_matching_loss": 0.022283492237329484, + "grad_norm": 0.22762863337993622, + "learning_rate": 2.7480993511056913e-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": 0.011530674807727338, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.005384406144730747, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.02844575624912977, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.022283492237329484, + "weighted_rkd_loss": 0.014222878124564886 + }, + { + "epoch": 1.965811965811966, + "flow_matching_loss": 0.021728577837347985, + "grad_norm": 0.22287198901176453, + "learning_rate": 2.7425692502862684e-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": 0.010578506160527468, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.004825728223659098, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.025982740707695483, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.021728577837347985, + "weighted_rkd_loss": 0.012991370353847741 + }, + { + "epoch": 1.982905982905983, + "flow_matching_loss": 0.02327656075358391, + "grad_norm": 0.23794780671596527, + "learning_rate": 2.7369847954624728e-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": 0.011695891991257668, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.005463259224779904, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.02885504309087992, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.02327656075358391, + "weighted_rkd_loss": 0.01442752154543996 + }, + { + "epoch": 2.0, + "flow_matching_loss": 0.024072754383087158, + "grad_norm": 0.3043615221977234, + "learning_rate": 2.731346230916448e-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": 0.012071775365620851, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.005929509131237865, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.03007306009531021, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.024072754383087158, + "weighted_rkd_loss": 0.015036530047655105 + }, + { + "epoch": 2.017094017094017, + "flow_matching_loss": 0.02227348256856203, + "grad_norm": 0.25443077087402344, + "learning_rate": 2.7256538032972746e-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": 0.0115864600520581, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.004511854983866215, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.027684775181114674, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.02227348256856203, + "weighted_rkd_loss": 0.013842387590557337 + }, + { + "epoch": 2.034188034188034, + "flow_matching_loss": 0.024202101118862628, + "grad_norm": 0.26608726382255554, + "learning_rate": 2.7199077616101773e-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": 0.012317976541817188, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.005561098386533558, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.0301970511674881, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.024202101118862628, + "weighted_rkd_loss": 0.01509852558374405 + }, + { + "epoch": 2.051282051282051, + "flow_matching_loss": 0.022172436024993657, + "grad_norm": 0.27349424362182617, + "learning_rate": 2.714108357205634e-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": 0.011763661075383425, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.005359243531711399, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.02888656537979841, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.022172436024993657, + "weighted_rkd_loss": 0.014443282689899205 + }, + { + "epoch": 2.0683760683760686, + "flow_matching_loss": 0.024315887689590455, + "grad_norm": 0.2344953566789627, + "learning_rate": 2.7082558437683803e-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": 0.012187405768781901, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.005494609219022095, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.02986942045390606, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.024315887689590455, + "weighted_rkd_loss": 0.01493471022695303 + }, + { + "epoch": 2.0854700854700856, + "flow_matching_loss": 0.023484759498387575, + "grad_norm": 0.221843883395195, + "learning_rate": 2.702350477306315e-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": 0.010893863625824451, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0050899856723845005, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.026877713017165662, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.023484759498387575, + "weighted_rkd_loss": 0.013438856508582831 + }, + { + "epoch": 2.1025641025641026, + "flow_matching_loss": 0.023697759956121445, + "grad_norm": 0.22658027708530426, + "learning_rate": 2.6963925161392963e-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": 0.011377280345186591, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.004846726846881211, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.027601287700235844, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.023697759956121445, + "weighted_rkd_loss": 0.013800643850117922 + }, + { + "epoch": 2.1196581196581197, + "flow_matching_loss": 0.02271516965702176, + "grad_norm": 0.2650354206562042, + "learning_rate": 2.6903822208878476e-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": 0.011314871022477746, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.005111789773218333, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.02774153146892786, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.02271516965702176, + "weighted_rkd_loss": 0.01387076573446393 + }, + { + "epoch": 2.1367521367521367, + "flow_matching_loss": 0.01972265476360917, + "grad_norm": 0.22902297973632812, + "learning_rate": 2.6843198544617535e-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": 0.011345456866547466, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.004953816230408847, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.02764473035931587, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.01972265476360917, + "weighted_rkd_loss": 0.013822365179657936 + }, + { + "epoch": 2.1538461538461537, + "flow_matching_loss": 0.023779618181288244, + "grad_norm": 0.20892459154129028, + "learning_rate": 2.6782056820485612e-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": 0.011763712484389543, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0052859523799270395, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.02881337758153677, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.023779618181288244, + "weighted_rkd_loss": 0.014406688790768386 + }, + { + "epoch": 2.1709401709401708, + "flow_matching_loss": 0.024139446392655372, + "grad_norm": 0.25326645374298096, + "learning_rate": 2.672039971101979e-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": 0.012011226126924157, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.005220527551136911, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.029242980107665063, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.024139446392655372, + "weighted_rkd_loss": 0.014621490053832531 + }, + { + "epoch": 2.1880341880341883, + "flow_matching_loss": 0.022985098883509636, + "grad_norm": 0.24101075530052185, + "learning_rate": 2.665822991330178e-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": 0.011152001190930605, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0050948970252648, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.02739889957010746, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.022985098883509636, + "weighted_rkd_loss": 0.01369944978505373 + }, + { + "epoch": 2.2051282051282053, + "flow_matching_loss": 0.022544006071984767, + "grad_norm": 0.26999637484550476, + "learning_rate": 2.6595550146839934e-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": 0.010217663925141096, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.004590991255827248, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.025026319175958635, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.022544006071984767, + "weighted_rkd_loss": 0.012513159587979317 + }, + { + "epoch": 2.2222222222222223, + "flow_matching_loss": 0.023427398689091206, + "grad_norm": 0.24688409268856049, + "learning_rate": 2.6532363153450287e-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": 0.011868452653288841, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0051475669723004104, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.02888447204604745, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.023427398689091206, + "weighted_rkd_loss": 0.014442236023023724 + }, + { + "epoch": 2.2393162393162394, + "flow_matching_loss": 0.018986200354993342, + "grad_norm": 0.1974107176065445, + "learning_rate": 2.6468671697136634e-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": 0.010647391248494387, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.004695117240771651, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.025989899411797523, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.018986200354993342, + "weighted_rkd_loss": 0.012994949705898762 + }, + { + "epoch": 2.2564102564102564, + "flow_matching_loss": 0.020662099216133357, + "grad_norm": 0.25234344601631165, + "learning_rate": 2.640447856396961e-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": 0.01112338388338685, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.004726927471347153, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.026973695401102303, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.020662099216133357, + "weighted_rkd_loss": 0.013486847700551152 + }, + { + "epoch": 2.2735042735042734, + "flow_matching_loss": 0.02161170719191432, + "grad_norm": 0.26437777280807495, + "learning_rate": 2.6339786561964805e-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": 0.011682726396247744, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0053446535021066666, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.0287101062014699, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.02161170719191432, + "weighted_rkd_loss": 0.01435505310073495 + }, + { + "epoch": 2.2905982905982905, + "flow_matching_loss": 0.019854203797876835, + "grad_norm": 0.27787142992019653, + "learning_rate": 2.6274598520959966e-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": 0.010752155911177397, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.004988197446800768, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.02649250943213701, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.019854203797876835, + "weighted_rkd_loss": 0.013246254716068506 + }, + { + "epoch": 2.3076923076923075, + "flow_matching_loss": 0.019564719311892985, + "grad_norm": 0.22497741878032684, + "learning_rate": 2.6208917292491185e-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": 0.010533380694687367, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.00482189361937344, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.025888654962182046, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.019564719311892985, + "weighted_rkd_loss": 0.012944327481091023 + }, + { + "epoch": 2.324786324786325, + "flow_matching_loss": 0.01939794821664691, + "grad_norm": 0.15387502312660217, + "learning_rate": 2.6142745749668176e-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": 0.009907269338145851, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.004406229685992002, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.024220768455415966, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.01939794821664691, + "weighted_rkd_loss": 0.012110384227707983 + }, + { + "epoch": 2.341880341880342, + "flow_matching_loss": 0.01964336708188057, + "grad_norm": 0.2878411114215851, + "learning_rate": 2.607608678704859e-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": 0.009940327517688275, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.004432538198307157, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.024313193187117575, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.01964336708188057, + "weighted_rkd_loss": 0.012156596593558788 + }, + { + "epoch": 2.358974358974359, + "flow_matching_loss": 0.02152224536985159, + "grad_norm": 0.23696327209472656, + "learning_rate": 2.600894332051139e-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": 0.011277066823095084, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.004696856765076518, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.027250990830361844, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.02152224536985159, + "weighted_rkd_loss": 0.013625495415180922 + }, + { + "epoch": 2.376068376068376, + "flow_matching_loss": 0.023002980556339026, + "grad_norm": 0.1936395913362503, + "learning_rate": 2.594131828712932e-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": 0.01043511051684618, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.004547051596455276, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.025417272560298444, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.023002980556339026, + "weighted_rkd_loss": 0.012708636280149222 + }, + { + "epoch": 2.393162393162393, + "flow_matching_loss": 0.020824392139911652, + "grad_norm": 0.18542948365211487, + "learning_rate": 2.5873214645040414e-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": 0.010291379503905773, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0044552497798576954, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.025038009230047464, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.020824392139911652, + "weighted_rkd_loss": 0.012519004615023732 + }, + { + "epoch": 2.41025641025641, + "flow_matching_loss": 0.022743955440819263, + "grad_norm": 0.24131953716278076, + "learning_rate": 2.5804635373318604e-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": 0.008490931289270521, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0036206896184012294, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.020602552220225334, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.022743955440819263, + "weighted_rkd_loss": 0.010301276110112667 + }, + { + "epoch": 2.427350427350427, + "flow_matching_loss": 0.020951343514025212, + "grad_norm": 0.2531515061855316, + "learning_rate": 2.57355834718434e-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": 0.01117038745433092, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.004377112933434546, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.026717887725681067, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.020951343514025212, + "weighted_rkd_loss": 0.013358943862840533 + }, + { + "epoch": 2.4444444444444446, + "flow_matching_loss": 0.022033390123397112, + "grad_norm": 0.23724442720413208, + "learning_rate": 2.566606196116867e-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": 0.012430550111457706, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.005375857185572386, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.030236957408487797, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.022033390123397112, + "weighted_rkd_loss": 0.015118478704243899 + }, + { + "epoch": 2.4615384615384617, + "flow_matching_loss": 0.022512477450072767, + "grad_norm": 0.25260433554649353, + "learning_rate": 2.559607388239051e-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": 0.011397971119731665, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.00474023618735373, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.027536178193986415, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.022512477450072767, + "weighted_rkd_loss": 0.013768089096993208 + }, + { + "epoch": 2.4786324786324787, + "flow_matching_loss": 0.018305157870054246, + "grad_norm": 0.25652942061424255, + "learning_rate": 2.552562229701422e-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": 0.010473096836358308, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.004887449718080461, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.02583364341408014, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.018305157870054246, + "weighted_rkd_loss": 0.01291682170704007 + }, + { + "epoch": 2.4957264957264957, + "flow_matching_loss": 0.02038935124874115, + "grad_norm": 0.19653008878231049, + "learning_rate": 2.545471028682036e-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": 0.008688374608755111, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0037490614922717214, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.02112581077963114, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.02038935124874115, + "weighted_rkd_loss": 0.01056290538981557 + }, + { + "epoch": 2.5128205128205128, + "flow_matching_loss": 0.021321542374789716, + "grad_norm": 0.19842712581157684, + "learning_rate": 2.538334095373e-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": 0.010492454003542661, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.004535641148686409, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.025520549155771732, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.021321542374789716, + "weighted_rkd_loss": 0.012760274577885866 + }, + { + "epoch": 2.52991452991453, + "flow_matching_loss": 0.021159323491156103, + "grad_norm": 0.2685452699661255, + "learning_rate": 2.531151741966896e-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": 0.010507429949939251, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.00491785912308842, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.02593271853402257, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.021159323491156103, + "weighted_rkd_loss": 0.012966359267011285 + }, + { + "epoch": 2.547008547008547, + "flow_matching_loss": 0.021657270099967717, + "grad_norm": 0.2372424602508545, + "learning_rate": 2.5239242826431302e-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": 0.00994455609470606, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0038265094626694918, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.02371562160551548, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.021657270099967717, + "weighted_rkd_loss": 0.01185781080275774 + }, + { + "epoch": 2.564102564102564, + "flow_matching_loss": 0.023013392556458713, + "grad_norm": 0.29201146960258484, + "learning_rate": 2.5166520335541865e-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": 0.01070014936849475, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.004663033946417272, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.026063332706689833, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.023013392556458713, + "weighted_rkd_loss": 0.013031666353344917 + }, + { + "epoch": 2.5811965811965814, + "flow_matching_loss": 0.021594572998583316, + "grad_norm": 0.3026672899723053, + "learning_rate": 2.5093353128117996e-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": 0.011556140519678593, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.005405857460573316, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.0285181388258934, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.021594572998583316, + "weighted_rkd_loss": 0.0142590694129467 + }, + { + "epoch": 2.5982905982905984, + "flow_matching_loss": 0.020032574795186518, + "grad_norm": 0.22828036546707153, + "learning_rate": 2.501974440473037e-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": 0.009677804866805672, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.004287159186787903, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.023642768524587153, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.020032574795186518, + "weighted_rkd_loss": 0.011821384262293577 + }, + { + "epoch": 2.6153846153846154, + "flow_matching_loss": 0.024566886201500892, + "grad_norm": 0.2961016595363617, + "learning_rate": 2.4945697385263015e-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": 0.010982732195407153, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.005046195024624467, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.02701165955513716, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.024566886201500892, + "weighted_rkd_loss": 0.01350582977756858 + }, + { + "epoch": 2.6324786324786325, + "flow_matching_loss": 0.01855727834627032, + "grad_norm": 0.17982171475887299, + "learning_rate": 2.487121530877243e-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": 0.009915598202496767, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.004172823298722506, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.02400401933118701, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.01855727834627032, + "weighted_rkd_loss": 0.012002009665593505 + }, + { + "epoch": 2.6495726495726495, + "flow_matching_loss": 0.019266926031559706, + "grad_norm": 0.2653079330921173, + "learning_rate": 2.4796301433345943e-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": 0.008381283609196544, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0037676420295611026, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.020530209038406612, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.019266926031559706, + "weighted_rkd_loss": 0.010265104519203306 + }, + { + "epoch": 2.6666666666666665, + "flow_matching_loss": 0.02113863080739975, + "grad_norm": 0.23463724553585052, + "learning_rate": 2.4720959035959148e-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": 0.011021296540275217, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.005067190737463534, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.02710978388786316, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.02113863080739975, + "weighted_rkd_loss": 0.01355489194393158 + }, + { + "epoch": 2.683760683760684, + "flow_matching_loss": 0.019188374001532792, + "grad_norm": 0.2564135491847992, + "learning_rate": 2.464519141233257e-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": 0.009046574961394072, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0036412632558494805, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.021734413225203753, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.019188374001532792, + "weighted_rkd_loss": 0.010867206612601876 + }, + { + "epoch": 2.700854700854701, + "flow_matching_loss": 0.020162386819720268, + "grad_norm": 0.23701418936252594, + "learning_rate": 2.456900187678754e-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": 0.008979757409542799, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0037617131136357786, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.021721228212118148, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.020162386819720268, + "weighted_rkd_loss": 0.010860614106059074 + }, + { + "epoch": 2.717948717948718, + "flow_matching_loss": 0.017920718621462585, + "grad_norm": 0.28698429465293884, + "learning_rate": 2.4492393762101147e-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": 0.009534297790378333, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0046347517287358645, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.023703347146511077, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.017920718621462585, + "weighted_rkd_loss": 0.011851673573255539 + }, + { + "epoch": 2.735042735042735, + "flow_matching_loss": 0.021741191670298577, + "grad_norm": 0.2914965748786926, + "learning_rate": 2.4415370419360508e-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": 0.009846937796100973, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.004321653652004898, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.02401552926748991, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.021741191670298577, + "weighted_rkd_loss": 0.012007764633744955 + }, + { + "epoch": 2.752136752136752, + "flow_matching_loss": 0.02044155988842249, + "grad_norm": 0.208698570728302, + "learning_rate": 2.4337935217816157e-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": 0.008389530563727021, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.003628258523531258, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.02040731953456998, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.02044155988842249, + "weighted_rkd_loss": 0.01020365976728499 + }, + { + "epoch": 2.769230769230769, + "flow_matching_loss": 0.023976971674710513, + "grad_norm": 0.3270052373409271, + "learning_rate": 2.4260091544734665e-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": 0.00906022354029119, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.004299577977508307, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.022420025151222944, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.023976971674710513, + "weighted_rkd_loss": 0.011210012575611472 + }, + { + "epoch": 2.786324786324786, + "flow_matching_loss": 0.020799010060727596, + "grad_norm": 0.24127914011478424, + "learning_rate": 2.4181842805250466e-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": 0.00902047767303884, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.00412689617369324, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.022167851496487857, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.020799010060727596, + "weighted_rkd_loss": 0.011083925748243928 + }, + { + "epoch": 2.8034188034188032, + "flow_matching_loss": 0.019217000994831323, + "grad_norm": 0.19288498163223267, + "learning_rate": 2.4103192422216923e-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": 0.008241457166150212, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0038245007395744323, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.020307415258139373, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.019217000994831323, + "weighted_rkd_loss": 0.010153707629069687 + }, + { + "epoch": 2.8205128205128203, + "flow_matching_loss": 0.02023093234747648, + "grad_norm": 0.23065443336963654, + "learning_rate": 2.402414383605658e-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": 0.01002774629741907, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.004230923461727798, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.02428641626611352, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.02023093234747648, + "weighted_rkd_loss": 0.01214320813305676 + }, + { + "epoch": 2.8376068376068377, + "flow_matching_loss": 0.019976893067359926, + "grad_norm": 0.23972664773464203, + "learning_rate": 2.3944700504610674e-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": 0.009053677832707763, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.003988711698912084, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.022096067015081645, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.019976893067359926, + "weighted_rkd_loss": 0.011048033507540822 + }, + { + "epoch": 2.8547008547008548, + "flow_matching_loss": 0.019993512891232967, + "grad_norm": 0.20631352066993713, + "learning_rate": 2.3864865902987883e-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": 0.011681730439886451, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.00535094968508929, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.02871441077440977, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.019993512891232967, + "weighted_rkd_loss": 0.014357205387204886 + }, + { + "epoch": 2.871794871794872, + "flow_matching_loss": 0.022632540948688985, + "grad_norm": 0.402326762676239, + "learning_rate": 2.378464352341231e-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": 0.009516790183261037, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.004438832867890596, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.023472413420677185, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.022632540948688985, + "weighted_rkd_loss": 0.011736206710338593 + }, + { + "epoch": 2.888888888888889, + "flow_matching_loss": 0.01911340495571494, + "grad_norm": 0.16892953217029572, + "learning_rate": 2.3704036875070728e-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": 0.0086509948130697, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0036977678537368774, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.020999757573008537, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.01911340495571494, + "weighted_rkd_loss": 0.010499878786504269 + }, + { + "epoch": 2.905982905982906, + "flow_matching_loss": 0.020744985435158014, + "grad_norm": 0.33316120505332947, + "learning_rate": 2.362304948395906e-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": 0.010707092797383666, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.004412930202670395, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.02582711586728692, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.020744985435158014, + "weighted_rkd_loss": 0.01291355793364346 + }, + { + "epoch": 2.9230769230769234, + "flow_matching_loss": 0.020460900105535983, + "grad_norm": 0.20060962438583374, + "learning_rate": 2.3541684892728157e-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": 0.009276452753692866, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.004033991321921349, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.02258689682930708, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.020460900105535983, + "weighted_rkd_loss": 0.01129344841465354 + }, + { + "epoch": 2.9401709401709404, + "flow_matching_loss": 0.02339051980525255, + "grad_norm": 0.25455862283706665, + "learning_rate": 2.345994666052883e-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": 0.009975053742527962, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0043854756280779835, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.024335582740604877, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.02339051980525255, + "weighted_rkd_loss": 0.012167791370302439 + }, + { + "epoch": 2.9572649572649574, + "flow_matching_loss": 0.018615886382758618, + "grad_norm": 0.19881266355514526, + "learning_rate": 2.337783836285614e-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": 0.008039046730846166, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0035203792504034936, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.019598472397774458, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.018615886382758618, + "weighted_rkd_loss": 0.009799236198887229 + }, + { + "epoch": 2.9743589743589745, + "flow_matching_loss": 0.017816959600895643, + "grad_norm": 0.24867010116577148, + "learning_rate": 2.3295363591393026e-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": 0.008814692776650191, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0037556086434051396, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.021384994313120843, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.017816959600895643, + "weighted_rkd_loss": 0.010692497156560421 + }, + { + "epoch": 2.9914529914529915, + "flow_matching_loss": 0.019321411941200496, + "grad_norm": 0.17765365540981293, + "learning_rate": 2.3212525953853174e-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": 0.008404076425358653, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0034041425213217737, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.020212295465171337, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.019321411941200496, + "weighted_rkd_loss": 0.010106147732585669 + }, + { + "epoch": 3.0085470085470085, + "flow_matching_loss": 0.01773095056414604, + "grad_norm": 0.21607407927513123, + "learning_rate": 2.3129329073823206e-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": 0.00853626844473183, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0038150409469380974, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.020887577999383213, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.01773095056414604, + "weighted_rkd_loss": 0.010443788999691606 + }, + { + "epoch": 3.0256410256410255, + "flow_matching_loss": 0.02112242393195629, + "grad_norm": 0.18876363337039948, + "learning_rate": 2.3045776590604175e-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": 0.008076513698324562, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0036738408263772724, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.019826868269592523, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.02112242393195629, + "weighted_rkd_loss": 0.009913434134796261 + }, + { + "epoch": 3.0427350427350426, + "flow_matching_loss": 0.01748533984646201, + "grad_norm": 0.20786572992801666, + "learning_rate": 2.296187215905237e-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": 0.00887304889038205, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.003883872437290847, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.021629970334470273, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.01748533984646201, + "weighted_rkd_loss": 0.010814985167235136 + }, + { + "epoch": 3.0598290598290596, + "flow_matching_loss": 0.019942130148410796, + "grad_norm": 0.2008289098739624, + "learning_rate": 2.2877619449419437e-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": 0.007430748268961907, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.003051865240558982, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.017913361731916665, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.019942130148410796, + "weighted_rkd_loss": 0.008956680865958333 + }, + { + "epoch": 3.076923076923077, + "flow_matching_loss": 0.017858355678617953, + "grad_norm": 0.23648367822170258, + "learning_rate": 2.2793022147191843e-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": 0.008725933264940977, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0039275804534554485, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.021379446797072887, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.017858355678617953, + "weighted_rkd_loss": 0.010689723398536443 + }, + { + "epoch": 3.094017094017094, + "flow_matching_loss": 0.019460619427263737, + "grad_norm": 0.19446852803230286, + "learning_rate": 2.2708083952929634e-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": 0.00935962744988501, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.004221779410727322, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.022941034566611052, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.019460619427263737, + "weighted_rkd_loss": 0.011470517283305526 + }, + { + "epoch": 3.111111111111111, + "flow_matching_loss": 0.020608755759894848, + "grad_norm": 0.1771230548620224, + "learning_rate": 2.2622808582104594e-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": 0.008036327594891191, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0034432843327522277, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01951593942940235, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.020608755759894848, + "weighted_rkd_loss": 0.009757969714701175 + }, + { + "epoch": 3.128205128205128, + "flow_matching_loss": 0.019297568406909705, + "grad_norm": 0.21222223341464996, + "learning_rate": 2.2537199764937703e-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": 0.008185308333486319, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.003753257519565523, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.020123874116688967, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.019297568406909705, + "weighted_rkd_loss": 0.010061937058344484 + }, + { + "epoch": 3.1452991452991452, + "flow_matching_loss": 0.01877512726932764, + "grad_norm": 0.22150124609470367, + "learning_rate": 2.245126124623595e-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": 0.007461652671918273, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0030335982097312807, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.017956903483718634, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.01877512726932764, + "weighted_rkd_loss": 0.008978451741859317 + }, + { + "epoch": 3.1623931623931623, + "flow_matching_loss": 0.019158876594156028, + "grad_norm": 0.24881495535373688, + "learning_rate": 2.2364996785228547e-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": 0.008031777897849679, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0036358975106850267, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.019699453469365837, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.019158876594156028, + "weighted_rkd_loss": 0.009849726734682918 + }, + { + "epoch": 3.1794871794871793, + "flow_matching_loss": 0.019298009760677814, + "grad_norm": 0.22011765837669373, + "learning_rate": 2.227841015540248e-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": 0.007320525776594877, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0032845957670360804, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.017925647273659705, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.019298009760677814, + "weighted_rkd_loss": 0.008962823636829853 + }, + { + "epoch": 3.1965811965811968, + "flow_matching_loss": 0.018039000034332276, + "grad_norm": 0.19180837273597717, + "learning_rate": 2.2191505144337433e-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": 0.00708873039111495, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0030891754431650043, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.017266636062413455, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.018039000034332276, + "weighted_rkd_loss": 0.008633318031206727 + }, + { + "epoch": 3.213675213675214, + "flow_matching_loss": 0.019442345015704632, + "grad_norm": 0.18389953672885895, + "learning_rate": 2.210428555354013e-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": 0.008321293070912361, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.003208835725672543, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01985142193734646, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.019442345015704632, + "weighted_rkd_loss": 0.00992571096867323 + }, + { + "epoch": 3.230769230769231, + "flow_matching_loss": 0.018029146548360585, + "grad_norm": 0.22258521616458893, + "learning_rate": 2.201675519827802e-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": 0.009321912564337254, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.004251653968822211, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.02289547910913825, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.018029146548360585, + "weighted_rkd_loss": 0.011447739554569125 + }, + { + "epoch": 3.247863247863248, + "flow_matching_loss": 0.019070112984627484, + "grad_norm": 0.19786061346530914, + "learning_rate": 2.1928917907412403e-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": 0.006793894432485103, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0031313693965785206, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.016719158180058, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.019070112984627484, + "weighted_rkd_loss": 0.008359579090029 + }, + { + "epoch": 3.264957264957265, + "flow_matching_loss": 0.019575916323810817, + "grad_norm": 0.22130337357521057, + "learning_rate": 2.1840777523230937e-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": 0.007430389663204551, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0033665582071989776, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01822733711451292, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.019575916323810817, + "weighted_rkd_loss": 0.00911366855725646 + }, + { + "epoch": 3.282051282051282, + "flow_matching_loss": 0.019690101593732835, + "grad_norm": 0.19864045083522797, + "learning_rate": 2.1752337901279552e-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": 0.01041108025237918, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.004671746958047151, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.025493907742202282, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.019690101593732835, + "weighted_rkd_loss": 0.012746953871101141 + }, + { + "epoch": 3.299145299145299, + "flow_matching_loss": 0.017883980367332696, + "grad_norm": 0.2654552459716797, + "learning_rate": 2.166360291019383e-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": 0.007775595178827643, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0032236144063062964, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.018774804752320053, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.017883980367332696, + "weighted_rkd_loss": 0.009387402376160026 + }, + { + "epoch": 3.316239316239316, + "flow_matching_loss": 0.019346895068883894, + "grad_norm": 0.2042943388223648, + "learning_rate": 2.157457643152973e-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": 0.008372717257589101, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.003984238905832172, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.020729673374444247, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.019346895068883894, + "weighted_rkd_loss": 0.010364836687222123 + }, + { + "epoch": 3.3333333333333335, + "flow_matching_loss": 0.019471426773816348, + "grad_norm": 0.25357481837272644, + "learning_rate": 2.148526235959385e-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": 0.008183046244084835, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0032328652683645486, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.019598957616835833, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.019471426773816348, + "weighted_rkd_loss": 0.009799478808417916 + }, + { + "epoch": 3.3504273504273505, + "flow_matching_loss": 0.01991074988618493, + "grad_norm": 0.34914514422416687, + "learning_rate": 2.1395664601273016e-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": 0.009180402988567949, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0043913544155657295, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.02275216057896614, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.01991074988618493, + "weighted_rkd_loss": 0.01137608028948307 + }, + { + "epoch": 3.3675213675213675, + "flow_matching_loss": 0.01942356275394559, + "grad_norm": 0.2123727649450302, + "learning_rate": 2.130578707586345e-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": 0.007257910165935755, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.003356795373838395, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.017872615810483695, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.01942356275394559, + "weighted_rkd_loss": 0.008936307905241848 + }, + { + "epoch": 3.3846153846153846, + "flow_matching_loss": 0.020389499142766, + "grad_norm": 0.20266097784042358, + "learning_rate": 2.1215633714899264e-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": 0.008410029159858824, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.003946584858931601, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.020766643062233924, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.020389499142766, + "weighted_rkd_loss": 0.010383321531116962 + }, + { + "epoch": 3.4017094017094016, + "flow_matching_loss": 0.019058057852089406, + "grad_norm": 0.19904877245426178, + "learning_rate": 2.112520846198053e-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": 0.007345339097082615, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.003269709087908268, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.017960387282073498, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.019058057852089406, + "weighted_rkd_loss": 0.008980193641036749 + }, + { + "epoch": 3.4188034188034186, + "flow_matching_loss": 0.01751188179478049, + "grad_norm": 0.23957492411136627, + "learning_rate": 2.103451527260074e-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": 0.009770218981429935, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.003764192934613675, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.023304630909115075, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.01751188179478049, + "weighted_rkd_loss": 0.011652315454557537 + }, + { + "epoch": 3.435897435897436, + "flow_matching_loss": 0.01685391329228878, + "grad_norm": 0.2283620834350586, + "learning_rate": 2.094355811397381e-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": 0.008938308944925666, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0036955745657905937, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.021572192572057246, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.01685391329228878, + "weighted_rkd_loss": 0.010786096286028623 + }, + { + "epoch": 3.452991452991453, + "flow_matching_loss": 0.020927964802831413, + "grad_norm": 0.19366750121116638, + "learning_rate": 2.0852340964860522e-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": 0.007808802416548133, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0034792908933013677, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.019096895679831506, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.020927964802831413, + "weighted_rkd_loss": 0.009548447839915753 + }, + { + "epoch": 3.47008547008547, + "flow_matching_loss": 0.01962413191795349, + "grad_norm": 0.2667924463748932, + "learning_rate": 2.076086781539447e-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": 0.008146571135148406, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0036341695114970205, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.019927312061190604, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.01962413191795349, + "weighted_rkd_loss": 0.009963656030595302 + }, + { + "epoch": 3.4871794871794872, + "flow_matching_loss": 0.01691184686496854, + "grad_norm": 0.2451614886522293, + "learning_rate": 2.066914266690755e-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": 0.00778640741482377, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0034134221728891133, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.018986237235367298, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.01691184686496854, + "weighted_rkd_loss": 0.009493118617683649 + }, + { + "epoch": 3.5042735042735043, + "flow_matching_loss": 0.018288571666926146, + "grad_norm": 0.17892566323280334, + "learning_rate": 2.0577169531754906e-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": 0.009745646128430962, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.004105433961376548, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.02359672598540783, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.018288571666926146, + "weighted_rkd_loss": 0.011798362992703915 + }, + { + "epoch": 3.5213675213675213, + "flow_matching_loss": 0.019396532233804463, + "grad_norm": 0.2194850742816925, + "learning_rate": 2.048495243313942e-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": 0.00801478703506291, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.003183236357290298, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.019212810415774585, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.019396532233804463, + "weighted_rkd_loss": 0.009606405207887293 + }, + { + "epoch": 3.5384615384615383, + "flow_matching_loss": 0.016854341328144073, + "grad_norm": 0.25259578227996826, + "learning_rate": 2.039249540493572e-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": 0.008965770015493035, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.003583391779102385, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.021514931693673134, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.016854341328144073, + "weighted_rkd_loss": 0.010757465846836567 + }, + { + "epoch": 3.5555555555555554, + "flow_matching_loss": 0.017057746276259424, + "grad_norm": 0.24107228219509125, + "learning_rate": 2.0299802491513755e-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": 0.00825087893754244, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0033832124434411527, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.019884970411658286, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.017057746276259424, + "weighted_rkd_loss": 0.009942485205829143 + }, + { + "epoch": 3.5726495726495724, + "flow_matching_loss": 0.017639929056167604, + "grad_norm": 0.22636933624744415, + "learning_rate": 2.0206877747561828e-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": 0.00801793122664094, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.003316805337090045, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.019352667685598134, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.017639929056167604, + "weighted_rkd_loss": 0.009676333842799067 + }, + { + "epoch": 3.58974358974359, + "flow_matching_loss": 0.017501342855393887, + "grad_norm": 0.2545315623283386, + "learning_rate": 2.011372523790927e-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": 0.0090520229190588, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.003841168456710875, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.02194521427154541, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.017501342855393887, + "weighted_rkd_loss": 0.010972607135772704 + }, + { + "epoch": 3.606837606837607, + "flow_matching_loss": 0.01844058623537421, + "grad_norm": 0.2210749238729477, + "learning_rate": 2.0020349037348643e-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": 0.0076704264618456365, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.003747639432549477, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.019088492263108492, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.01844058623537421, + "weighted_rkd_loss": 0.009544246131554246 + }, + { + "epoch": 3.623931623931624, + "flow_matching_loss": 0.017102108057588338, + "grad_norm": 0.2192651629447937, + "learning_rate": 1.9926753230457443e-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": 0.010010592406615615, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.004023256804794073, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.02404444171115756, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.017102108057588338, + "weighted_rkd_loss": 0.01202222085557878 + }, + { + "epoch": 3.641025641025641, + "flow_matching_loss": 0.02092074267566204, + "grad_norm": 0.19454717636108398, + "learning_rate": 1.983294191141948e-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": 0.007164185820147395, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0030857678735628723, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01741413939744234, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.02092074267566204, + "weighted_rkd_loss": 0.00870706969872117 + }, + { + "epoch": 3.658119658119658, + "flow_matching_loss": 0.019593456014990806, + "grad_norm": 0.21146218478679657, + "learning_rate": 1.973891918384575e-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": 0.0082511723972857, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0033616421744227408, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.019863986968994142, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.019593456014990806, + "weighted_rkd_loss": 0.009931993484497071 + }, + { + "epoch": 3.6752136752136755, + "flow_matching_loss": 0.020794244669377803, + "grad_norm": 0.26112446188926697, + "learning_rate": 1.9644689160594945e-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": 0.008243131591007114, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.003457737248390913, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.019944000616669656, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.020794244669377803, + "weighted_rkd_loss": 0.009972000308334828 + }, + { + "epoch": 3.6923076923076925, + "flow_matching_loss": 0.016974092926830053, + "grad_norm": 0.16520389914512634, + "learning_rate": 1.9550255963593553e-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": 0.006980267446488142, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.002796297729946673, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.016756832785904407, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.016974092926830053, + "weighted_rkd_loss": 0.008378416392952203 + }, + { + "epoch": 3.7094017094017095, + "flow_matching_loss": 0.016115081124007703, + "grad_norm": 0.2137187272310257, + "learning_rate": 1.9455623723655524e-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": 0.006464245729148388, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.002947816206142306, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.015876307524740695, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.016115081124007703, + "weighted_rkd_loss": 0.007938153762370347 + }, + { + "epoch": 3.7264957264957266, + "flow_matching_loss": 0.01735110590234399, + "grad_norm": 0.23739102482795715, + "learning_rate": 1.9360796580301606e-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": 0.00815049447119236, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.003459366550669074, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.019760355912148954, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.01735110590234399, + "weighted_rkd_loss": 0.009880177956074477 + }, + { + "epoch": 3.7435897435897436, + "flow_matching_loss": 0.01909026876091957, + "grad_norm": 0.20851047337055206, + "learning_rate": 1.9265778681578245e-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": 0.008637681510299445, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0034640667494386435, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.02073942981660366, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.01909026876091957, + "weighted_rkd_loss": 0.01036971490830183 + }, + { + "epoch": 3.7606837606837606, + "flow_matching_loss": 0.017638309486210348, + "grad_norm": 0.2060084491968155, + "learning_rate": 1.9170574183876143e-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": 0.006922317855060101, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0031611088663339613, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.017005744390189648, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.017638309486210348, + "weighted_rkd_loss": 0.008502872195094824 + }, + { + "epoch": 3.7777777777777777, + "flow_matching_loss": 0.01585266077890992, + "grad_norm": 0.23067165911197662, + "learning_rate": 1.9075187251748452e-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": 0.006699414411559701, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0027543757925741373, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01615320462733507, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.01585266077890992, + "weighted_rkd_loss": 0.008076602313667535 + }, + { + "epoch": 3.7948717948717947, + "flow_matching_loss": 0.016888183262199163, + "grad_norm": 0.25637298822402954, + "learning_rate": 1.8979622057728596e-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": 0.0072902919258922335, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0030389142571948468, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.017619498074054718, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.016888183262199163, + "weighted_rkd_loss": 0.008809749037027359 + }, + { + "epoch": 3.8119658119658117, + "flow_matching_loss": 0.019817265029996634, + "grad_norm": 0.180880606174469, + "learning_rate": 1.8883882782147748e-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": 0.00791191435419023, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.003189689852297306, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.019013518746942282, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.019817265029996634, + "weighted_rkd_loss": 0.009506759373471141 + }, + { + "epoch": 3.8290598290598292, + "flow_matching_loss": 0.018800769094377758, + "grad_norm": 0.2036091685295105, + "learning_rate": 1.8787973612951967e-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": 0.007398828910663724, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0029260202194564045, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.017723677959293128, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.018800769094377758, + "weighted_rkd_loss": 0.008861838979646564 + }, + { + "epoch": 3.8461538461538463, + "flow_matching_loss": 0.01811023661866784, + "grad_norm": 0.19988355040550232, + "learning_rate": 1.8691898745519028e-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": 0.007038753991946578, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0028219386003911495, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.016899446304887534, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.01811023661866784, + "weighted_rkd_loss": 0.008449723152443767 + }, + { + "epoch": 3.8632478632478633, + "flow_matching_loss": 0.01981883533298969, + "grad_norm": 0.2408934384584427, + "learning_rate": 1.8595662382474863e-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": 0.006067744246684015, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0026935314759612083, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01482901992276311, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.01981883533298969, + "weighted_rkd_loss": 0.007414509961381555 + }, + { + "epoch": 3.8803418803418803, + "flow_matching_loss": 0.017147166561335324, + "grad_norm": 0.22736790776252747, + "learning_rate": 1.8499268733509766e-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": 0.007045287149958312, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0030606971587985752, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01715127155184746, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.017147166561335324, + "weighted_rkd_loss": 0.00857563577592373 + }, + { + "epoch": 3.8974358974358974, + "flow_matching_loss": 0.019914013613015414, + "grad_norm": 0.18116098642349243, + "learning_rate": 1.84027220151942e-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": 0.007360791740939021, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.003151774185243994, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01787335779517889, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.019914013613015414, + "weighted_rkd_loss": 0.008936678897589445 + }, + { + "epoch": 3.9145299145299144, + "flow_matching_loss": 0.01781042041257024, + "grad_norm": 0.2462361752986908, + "learning_rate": 1.83060264507944e-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": 0.006698027206584811, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0028518905979581176, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01624794499948621, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.01781042041257024, + "weighted_rkd_loss": 0.008123972499743105 + }, + { + "epoch": 3.931623931623932, + "flow_matching_loss": 0.0174762106500566, + "grad_norm": 0.2150101363658905, + "learning_rate": 1.82091862700876e-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": 0.00740150585770607, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0028824614943005144, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.017685473151504993, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.0174762106500566, + "weighted_rkd_loss": 0.008842736575752496 + }, + { + "epoch": 3.948717948717949, + "flow_matching_loss": 0.018270046450197697, + "grad_norm": 0.18908220529556274, + "learning_rate": 1.8112205709177023e-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": 0.007680607819929719, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0030518362997099756, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.018413052149116992, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.018270046450197697, + "weighted_rkd_loss": 0.009206526074558496 + }, + { + "epoch": 3.965811965811966, + "flow_matching_loss": 0.018343500047922134, + "grad_norm": 0.2890482544898987, + "learning_rate": 1.8015089010306585e-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": 0.007763550756499172, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.003291046607773751, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01881814803928137, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.018343500047922134, + "weighted_rkd_loss": 0.009409074019640685 + }, + { + "epoch": 3.982905982905983, + "flow_matching_loss": 0.01808697897940874, + "grad_norm": 0.24271777272224426, + "learning_rate": 1.7917840421675298e-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": 0.008296487387269735, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0034420376061461864, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.02003501234576106, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.01808697897940874, + "weighted_rkd_loss": 0.01001750617288053 + }, + { + "epoch": 4.0, + "flow_matching_loss": 0.01567900711670518, + "grad_norm": 0.18737336993217468, + "learning_rate": 1.7820464197251474e-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": 0.006013613659888506, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.002602348988875747, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.014629576355218887, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.01567900711670518, + "weighted_rkd_loss": 0.007314788177609444 + }, + { + "epoch": 4.017094017094017, + "flow_matching_loss": 0.021943020261824132, + "grad_norm": 0.20966759324073792, + "learning_rate": 1.7722964596586627e-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": 0.0065858547110110525, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.002628158254083246, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.015799867734313012, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.021943020261824132, + "weighted_rkd_loss": 0.007899933867156506 + }, + { + "epoch": 4.034188034188034, + "flow_matching_loss": 0.016713260672986507, + "grad_norm": 0.1769791692495346, + "learning_rate": 1.7625345884629144e-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": 0.006801077583804727, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0028325610561296343, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01643471634015441, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.016713260672986507, + "weighted_rkd_loss": 0.008217358170077205 + }, + { + "epoch": 4.051282051282051, + "flow_matching_loss": 0.019258282147347927, + "grad_norm": 0.15673449635505676, + "learning_rate": 1.7527612331537722e-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": 0.006572543457150459, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0027091597439721227, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01585424654185772, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.019258282147347927, + "weighted_rkd_loss": 0.00792712327092886 + }, + { + "epoch": 4.068376068376068, + "flow_matching_loss": 0.018021010048687457, + "grad_norm": 0.21731337904930115, + "learning_rate": 1.7429768212494593e-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": 0.0063932741759344935, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0027011008467525245, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01548764919862151, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.018021010048687457, + "weighted_rkd_loss": 0.007743824599310755 + }, + { + "epoch": 4.085470085470085, + "flow_matching_loss": 0.01657121330499649, + "grad_norm": 0.19339190423488617, + "learning_rate": 1.7331817807518497e-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": 0.006700467597693205, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0028941453667357562, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.016295080445706843, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.01657121330499649, + "weighted_rkd_loss": 0.008147540222853421 + }, + { + "epoch": 4.102564102564102, + "flow_matching_loss": 0.017072345502674578, + "grad_norm": 0.23046529293060303, + "learning_rate": 1.723376540127745e-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": 0.006289798207581044, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.002920631063170731, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.015500227455049753, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.017072345502674578, + "weighted_rkd_loss": 0.007750113727524877 + }, + { + "epoch": 4.119658119658119, + "flow_matching_loss": 0.017261643148958682, + "grad_norm": 0.2079930305480957, + "learning_rate": 1.7135615282901357e-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": 0.006375982658937573, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.002564786572474986, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.0153167518787086, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.017261643148958682, + "weighted_rkd_loss": 0.0076583759393543 + }, + { + "epoch": 4.136752136752137, + "flow_matching_loss": 0.01995796347036958, + "grad_norm": 0.1875106692314148, + "learning_rate": 1.7037371745794358e-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": 0.00826740562915802, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0032586575718596576, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.019793468713760375, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.01995796347036958, + "weighted_rkd_loss": 0.009896734356880188 + }, + { + "epoch": 4.153846153846154, + "flow_matching_loss": 0.01812446564435959, + "grad_norm": 0.13842140138149261, + "learning_rate": 1.6939039087447026e-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": 0.007443991838954389, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0030200185254216196, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.017908002063632012, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.01812446564435959, + "weighted_rkd_loss": 0.008954001031816006 + }, + { + "epoch": 4.170940170940171, + "flow_matching_loss": 0.017951147351413965, + "grad_norm": 0.2116854190826416, + "learning_rate": 1.6840621609248397e-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": 0.007201647013425827, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0029258668189868333, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.017329161055386066, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.017951147351413965, + "weighted_rkd_loss": 0.008664580527693033 + }, + { + "epoch": 4.188034188034188, + "flow_matching_loss": 0.01645336365327239, + "grad_norm": 0.19698773324489594, + "learning_rate": 1.6742123616297794e-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": 0.006453622179105878, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0028069419087842105, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.015714186150580643, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.01645336365327239, + "weighted_rkd_loss": 0.007857093075290322 + }, + { + "epoch": 4.205128205128205, + "flow_matching_loss": 0.01733739422634244, + "grad_norm": 0.16423113644123077, + "learning_rate": 1.664354941721651e-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": 0.0059458487667143345, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0024242091225460173, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.014315906632691621, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.01733739422634244, + "weighted_rkd_loss": 0.0071579533163458105 + }, + { + "epoch": 4.222222222222222, + "flow_matching_loss": 0.015578425768762828, + "grad_norm": 0.23557522892951965, + "learning_rate": 1.6544903323959355e-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": 0.007263116491958499, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0027161234989762305, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01724235648289323, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.015578425768762828, + "weighted_rkd_loss": 0.008621178241446615 + }, + { + "epoch": 4.239316239316239, + "flow_matching_loss": 0.01863284520804882, + "grad_norm": 0.2926975190639496, + "learning_rate": 1.6446189651626023e-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": 0.006898061465471983, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.002947836392559111, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01674395939335227, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.01863284520804882, + "weighted_rkd_loss": 0.008371979696676135 + }, + { + "epoch": 4.256410256410256, + "flow_matching_loss": 0.017656180914491414, + "grad_norm": 0.24675606191158295, + "learning_rate": 1.6347412718272316e-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": 0.006419377285055816, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0024581300443969668, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.015296884812414646, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.017656180914491414, + "weighted_rkd_loss": 0.007648442406207323 + }, + { + "epoch": 4.273504273504273, + "flow_matching_loss": 0.014918852131813765, + "grad_norm": 0.168844535946846, + "learning_rate": 1.624857684472131e-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": 0.007374631660059095, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0025617657345719637, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01731102904304862, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.014918852131813765, + "weighted_rkd_loss": 0.00865551452152431 + }, + { + "epoch": 4.2905982905982905, + "flow_matching_loss": 0.017407764215022326, + "grad_norm": 0.24161049723625183, + "learning_rate": 1.6149686354374283e-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": 0.006880494114011526, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0028089393395930528, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.016569927614182235, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.017407764215022326, + "weighted_rkd_loss": 0.008284963807091117 + }, + { + "epoch": 4.3076923076923075, + "flow_matching_loss": 0.01608648793771863, + "grad_norm": 0.17501430213451385, + "learning_rate": 1.605074557302165e-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": 0.00622005220502615, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.00257482387823984, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.015014928299933672, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.01608648793771863, + "weighted_rkd_loss": 0.007507464149966836 + }, + { + "epoch": 4.3247863247863245, + "flow_matching_loss": 0.019678328651934862, + "grad_norm": 0.17470689117908478, + "learning_rate": 1.59517588286537e-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": 0.006203674664720893, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0025529713602736592, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01496032066643238, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.019678328651934862, + "weighted_rkd_loss": 0.00748016033321619 + }, + { + "epoch": 4.3418803418803416, + "flow_matching_loss": 0.017429454531520606, + "grad_norm": 0.15037848055362701, + "learning_rate": 1.5852730451271324e-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": 0.006688070483505726, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0025841880589723585, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01596032902598381, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.017429454531520606, + "weighted_rkd_loss": 0.007980164512991906 + }, + { + "epoch": 4.358974358974359, + "flow_matching_loss": 0.01633572867140174, + "grad_norm": 0.18205097317695618, + "learning_rate": 1.5753664772696546e-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": 0.0061551171587780115, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0023771570296958087, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01468739123083651, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.01633572867140174, + "weighted_rkd_loss": 0.007343695615418255 + }, + { + "epoch": 4.3760683760683765, + "flow_matching_loss": 0.017297369055449964, + "grad_norm": 0.16873420774936676, + "learning_rate": 1.565456612638307e-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": 0.007353437272831798, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.003271087654866278, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.017977962270379066, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.017297369055449964, + "weighted_rkd_loss": 0.008988981135189533 + }, + { + "epoch": 4.3931623931623935, + "flow_matching_loss": 0.01667330777272582, + "grad_norm": 0.1701025515794754, + "learning_rate": 1.5555438847226732e-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": 0.005713472934439778, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0026690538972616194, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.014095999766141176, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.01667330777272582, + "weighted_rkd_loss": 0.007047999883070588 + }, + { + "epoch": 4.410256410256411, + "flow_matching_loss": 0.017146031092852353, + "grad_norm": 0.16161145269870758, + "learning_rate": 1.545628727137584e-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": 0.006327573349699378, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.002536358719225973, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.015191505383700132, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.017146031092852353, + "weighted_rkd_loss": 0.007595752691850066 + }, + { + "epoch": 4.427350427350428, + "flow_matching_loss": 0.016588612459599972, + "grad_norm": 0.15618957579135895, + "learning_rate": 1.535711573604153e-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": 0.006100028147920966, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.002751457190606743, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.014951513428241015, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.016588612459599972, + "weighted_rkd_loss": 0.007475756714120508 + }, + { + "epoch": 4.444444444444445, + "flow_matching_loss": 0.014824415929615498, + "grad_norm": 0.1938074678182602, + "learning_rate": 1.5257928579308032e-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": 0.005981717212125659, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.002613829378969967, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.014577263686805963, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.014824415929615498, + "weighted_rkd_loss": 0.007288631843402981 + }, + { + "epoch": 4.461538461538462, + "flow_matching_loss": 0.015400273352861404, + "grad_norm": 0.20121899247169495, + "learning_rate": 1.5158730139942902e-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": 0.006921102246269584, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0030046363011933864, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.016846840642392634, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.015400273352861404, + "weighted_rkd_loss": 0.008423420321196317 + }, + { + "epoch": 4.478632478632479, + "flow_matching_loss": 0.017446022760123015, + "grad_norm": 0.21146860718727112, + "learning_rate": 1.5059524757207239e-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": 0.006235840497538447, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.003034585400018841, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.015506266336888074, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.017446022760123015, + "weighted_rkd_loss": 0.007753133168444037 + }, + { + "epoch": 4.495726495726496, + "flow_matching_loss": 0.016702610068023205, + "grad_norm": 0.18468596041202545, + "learning_rate": 1.4960316770665866e-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": 0.005742358369752764, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.002364849718287587, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.013849566597491503, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.016702610068023205, + "weighted_rkd_loss": 0.0069247832987457516 + }, + { + "epoch": 4.512820512820513, + "flow_matching_loss": 0.016380259487777948, + "grad_norm": 0.17872101068496704, + "learning_rate": 1.4861110519997495e-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": 0.0053622016683220865, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0022134567378088834, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01293786009773612, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.016380259487777948, + "weighted_rkd_loss": 0.00646893004886806 + }, + { + "epoch": 4.52991452991453, + "flow_matching_loss": 0.016376509610563517, + "grad_norm": 0.2395770251750946, + "learning_rate": 1.4761910344804913e-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": 0.007237207470461726, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0029750985442660747, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.017449513357132674, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.016376509610563517, + "weighted_rkd_loss": 0.008724756678566337 + }, + { + "epoch": 4.547008547008547, + "flow_matching_loss": 0.018613293673843144, + "grad_norm": 0.18903741240501404, + "learning_rate": 1.466272058442516e-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": 0.006073881895281375, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0024457727326080204, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.014593536406755448, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.018613293673843144, + "weighted_rkd_loss": 0.007296768203377724 + }, + { + "epoch": 4.564102564102564, + "flow_matching_loss": 0.01539095789194107, + "grad_norm": 0.20536470413208008, + "learning_rate": 1.4563545577739676e-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": 0.006566722132265568, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.002762407041154802, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01589585142210126, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.01539095789194107, + "weighted_rkd_loss": 0.00794792571105063 + }, + { + "epoch": 4.581196581196581, + "flow_matching_loss": 0.01374195357784629, + "grad_norm": 0.2122846096754074, + "learning_rate": 1.4464389662984543e-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": 0.006492527201771736, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.002470004640053958, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.015455059334635734, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.01374195357784629, + "weighted_rkd_loss": 0.007727529667317867 + }, + { + "epoch": 4.598290598290598, + "flow_matching_loss": 0.013929103873670102, + "grad_norm": 0.23724037408828735, + "learning_rate": 1.4365257177560684e-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": 0.005504302121698857, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.002407100109849125, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.013415704481303692, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.013929103873670102, + "weighted_rkd_loss": 0.006707852240651846 + }, + { + "epoch": 4.615384615384615, + "flow_matching_loss": 0.017199409008026124, + "grad_norm": 0.14590302109718323, + "learning_rate": 1.426615245784416e-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": 0.006096316920593381, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0025702380342409016, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.014762871898710728, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.017199409008026124, + "weighted_rkd_loss": 0.007381435949355364 + }, + { + "epoch": 4.632478632478632, + "flow_matching_loss": 0.014162557199597359, + "grad_norm": 0.17976969480514526, + "learning_rate": 1.4167079838996468e-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": 0.005237145652063191, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0019188881386071443, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.012393179349601269, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.014162557199597359, + "weighted_rkd_loss": 0.006196589674800634 + }, + { + "epoch": 4.64957264957265, + "flow_matching_loss": 0.015029059536755086, + "grad_norm": 0.2356736809015274, + "learning_rate": 1.4068043654774897e-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": 0.006231556832790375, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.002463959448505193, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01492707310244441, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.015029059536755086, + "weighted_rkd_loss": 0.007463536551222205 + }, + { + "epoch": 4.666666666666667, + "flow_matching_loss": 0.015622436720877885, + "grad_norm": 0.1874004602432251, + "learning_rate": 1.3969048237342992e-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": 0.0062313829082995655, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0025877908221445976, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.015050556696951389, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.015622436720877885, + "weighted_rkd_loss": 0.007525278348475694 + }, + { + "epoch": 4.683760683760684, + "flow_matching_loss": 0.018168924003839494, + "grad_norm": 0.20384423434734344, + "learning_rate": 1.3870097917081011e-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": 0.005497690802440047, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0024658505455590783, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.013461232278496027, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.018168924003839494, + "weighted_rkd_loss": 0.006730616139248013 + }, + { + "epoch": 4.700854700854701, + "flow_matching_loss": 0.015590931847691536, + "grad_norm": 0.20453773438930511, + "learning_rate": 1.3771197022396508e-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": 0.006258559506386518, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.002934907143935561, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.015452026110142469, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.015590931847691536, + "weighted_rkd_loss": 0.007726013055071234 + }, + { + "epoch": 4.717948717948718, + "flow_matching_loss": 0.017632334027439356, + "grad_norm": 0.1687292456626892, + "learning_rate": 1.3672349879535015e-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": 0.007652779622003436, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.003389199625235051, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.018694758508354427, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.017632334027439356, + "weighted_rkd_loss": 0.009347379254177213 + }, + { + "epoch": 4.735042735042735, + "flow_matching_loss": 0.015019437856972217, + "grad_norm": 0.1987934559583664, + "learning_rate": 1.3573560812390781e-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": 0.005383615242317319, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0025623307446949184, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.013329561147838832, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.015019437856972217, + "weighted_rkd_loss": 0.006664780573919416 + }, + { + "epoch": 4.752136752136752, + "flow_matching_loss": 0.015525436028838158, + "grad_norm": 0.20620040595531464, + "learning_rate": 1.3474834142317635e-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": 0.005638892576098442, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0022671450045891107, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.013544930052012205, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.015525436028838158, + "weighted_rkd_loss": 0.006772465026006103 + }, + { + "epoch": 4.769230769230769, + "flow_matching_loss": 0.017463005054742096, + "grad_norm": 0.17975881695747375, + "learning_rate": 1.3376174187939955e-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": 0.005944303330034017, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.002467387414071709, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.014355993922799825, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.017463005054742096, + "weighted_rkd_loss": 0.007177996961399913 + }, + { + "epoch": 4.786324786324786, + "flow_matching_loss": 0.017058755829930307, + "grad_norm": 0.16552963852882385, + "learning_rate": 1.3277585264963761e-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": 0.00515214626211673, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0020511400187388064, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.012355432473123073, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.017058755829930307, + "weighted_rkd_loss": 0.006177716236561537 + }, + { + "epoch": 4.803418803418803, + "flow_matching_loss": 0.015618646889925003, + "grad_norm": 0.1786433458328247, + "learning_rate": 1.3179071685987928e-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": 0.00622903483454138, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0024549101828597484, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.0149129799567163, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.015618646889925003, + "weighted_rkd_loss": 0.00745648997835815 + }, + { + "epoch": 4.82051282051282, + "flow_matching_loss": 0.016987103130668403, + "grad_norm": 0.18596892058849335, + "learning_rate": 1.308063776031554e-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": 0.005815305956639349, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0019663846818730236, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.013596996618434786, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.016987103130668403, + "weighted_rkd_loss": 0.006798498309217393 + }, + { + "epoch": 4.837606837606837, + "flow_matching_loss": 0.01462665731087327, + "grad_norm": 0.18217001855373383, + "learning_rate": 1.2982287793765389e-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": 0.006055335560813546, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0022699115332216026, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.014380582701414823, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.01462665731087327, + "weighted_rkd_loss": 0.007190291350707412 + }, + { + "epoch": 4.854700854700854, + "flow_matching_loss": 0.01419622004032135, + "grad_norm": 0.16423466801643372, + "learning_rate": 1.2884026088483616e-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": 0.006006566388532519, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0027799732168205084, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.014793106028810144, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.01419622004032135, + "weighted_rkd_loss": 0.007396553014405072 + }, + { + "epoch": 4.871794871794872, + "flow_matching_loss": 0.018762329872697592, + "grad_norm": 0.1827438473701477, + "learning_rate": 1.2785856942755543e-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": 0.005305541632696986, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0022030977881513534, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.012814180972054601, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.018762329872697592, + "weighted_rkd_loss": 0.006407090486027301 + }, + { + "epoch": 4.888888888888889, + "flow_matching_loss": 0.016695990692824125, + "grad_norm": 0.14952440559864044, + "learning_rate": 1.2687784650817616e-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": 0.005178940435871482, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0018734027398750186, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.012231283448636531, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.016695990692824125, + "weighted_rkd_loss": 0.006115641724318266 + }, + { + "epoch": 4.905982905982906, + "flow_matching_loss": 0.017721740435808896, + "grad_norm": 0.2071308046579361, + "learning_rate": 1.2589813502669603e-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": 0.006604563258588314, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.002799827535636723, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.016008953843265773, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.017721740435808896, + "weighted_rkd_loss": 0.008004476921632886 + }, + { + "epoch": 4.923076923076923, + "flow_matching_loss": 0.01568200821056962, + "grad_norm": 0.1882401555776596, + "learning_rate": 1.2491947783886907e-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": 0.005879974691197276, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.002689003897830844, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.014448953233659268, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.01568200821056962, + "weighted_rkd_loss": 0.007224476616829634 + }, + { + "epoch": 4.94017094017094, + "flow_matching_loss": 0.01636640205979347, + "grad_norm": 0.2302120476961136, + "learning_rate": 1.2394191775433106e-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": 0.005514456843957305, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.002103069971781224, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.013131983624771238, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.01636640205979347, + "weighted_rkd_loss": 0.006565991812385619 + }, + { + "epoch": 4.957264957264957, + "flow_matching_loss": 0.01687027271836996, + "grad_norm": 0.23090745508670807, + "learning_rate": 1.2296549753472696e-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": 0.00608045062981546, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0022834445582702754, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.014444345701485872, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.01687027271836996, + "weighted_rkd_loss": 0.007222172850742936 + }, + { + "epoch": 4.9743589743589745, + "flow_matching_loss": 0.01453867731615901, + "grad_norm": 0.15958406031131744, + "learning_rate": 1.2199025989184032e-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": 0.005980615597218275, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0023445994476787744, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.014305830746889115, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.01453867731615901, + "weighted_rkd_loss": 0.0071529153734445575 + }, + { + "epoch": 4.9914529914529915, + "flow_matching_loss": 0.014788433164358138, + "grad_norm": 0.18608048558235168, + "learning_rate": 1.21016247485725e-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": 0.005408674338832497, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0027271160739473998, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.013544464670121669, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.014788433164358138, + "weighted_rkd_loss": 0.0067722323350608345 + }, + { + "epoch": 5.0085470085470085, + "flow_matching_loss": 0.01518021747469902, + "grad_norm": 0.16777537763118744, + "learning_rate": 1.2004350292283897e-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": 0.005128721287474036, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0019198581692762672, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.012177300779148937, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.01518021747469902, + "weighted_rkd_loss": 0.006088650389574468 + }, + { + "epoch": 5.0256410256410255, + "flow_matching_loss": 0.013603176549077034, + "grad_norm": 0.16358694434165955, + "learning_rate": 1.1907206875418068e-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": 0.005439571966417134, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0021702540456317366, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.013049397990107537, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.013603176549077034, + "weighted_rkd_loss": 0.006524698995053768 + }, + { + "epoch": 5.042735042735043, + "flow_matching_loss": 0.01461758343502879, + "grad_norm": 0.16626055538654327, + "learning_rate": 1.1810198747342769e-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": 0.00551510548684746, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.002374007261823863, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.013404218293726445, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.01461758343502879, + "weighted_rkd_loss": 0.0067021091468632225 + }, + { + "epoch": 5.05982905982906, + "flow_matching_loss": 0.01351362131536007, + "grad_norm": 0.15272045135498047, + "learning_rate": 1.1713330151507786e-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": 0.005524679366499185, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.002252029615920037, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01330138836055994, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.01351362131536007, + "weighted_rkd_loss": 0.00665069418027997 + }, + { + "epoch": 5.076923076923077, + "flow_matching_loss": 0.015635596960783003, + "grad_norm": 0.19087515771389008, + "learning_rate": 1.161660532525931e-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": 0.006162820104509592, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0028224783949553967, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01514811860397458, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.015635596960783003, + "weighted_rkd_loss": 0.00757405930198729 + }, + { + "epoch": 5.094017094017094, + "flow_matching_loss": 0.013611517194658519, + "grad_norm": 0.19723477959632874, + "learning_rate": 1.1520028499654584e-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": 0.005134547990746796, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0022663619834929704, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.012535457825288176, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.013611517194658519, + "weighted_rkd_loss": 0.006267728912644088 + }, + { + "epoch": 5.111111111111111, + "flow_matching_loss": 0.014483093842864036, + "grad_norm": 0.1699741780757904, + "learning_rate": 1.1423603899276833e-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": 0.005278613604605198, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0021448526182211935, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.012702079955488443, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.014483093842864036, + "weighted_rkd_loss": 0.006351039977744222 + }, + { + "epoch": 5.128205128205128, + "flow_matching_loss": 0.015141089726239444, + "grad_norm": 0.20546045899391174, + "learning_rate": 1.1327335742050442e-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": 0.005539287813007832, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.002340341580566019, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.013418917311355472, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.015141089726239444, + "weighted_rkd_loss": 0.006709458655677736 + } + ], + "logging_steps": 10, + "max_steps": 5000, + "num_input_tokens_seen": 0, + "num_train_epochs": 9, + "save_steps": 5001, + "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": 64, + "trial_name": null, + "trial_params": null +} diff --git a/FewShot/39tasks/5shot/FromRKD-Timewarp_FullactionHead/seed2/default/train_5shot_seed2_fullactionhead_5k/experiment_cfg/metadata.json b/FewShot/39tasks/5shot/FromRKD-Timewarp_FullactionHead/seed2/default/train_5shot_seed2_fullactionhead_5k/experiment_cfg/metadata.json new file mode 100644 index 0000000000000000000000000000000000000000..23dcfa871a88e56280f945357aef0296bd071711 --- /dev/null +++ b/FewShot/39tasks/5shot/FromRKD-Timewarp_FullactionHead/seed2/default/train_5shot_seed2_fullactionhead_5k/experiment_cfg/metadata.json @@ -0,0 +1,431 @@ +{ + "new_embodiment": { + "statistics": { + "state": { + "base_position": { + "max": [ + 7.778976917266846, + 0.33403685688972473, + 0.7223060727119446 + ], + "min": [ + -4.821934223175049, + -6.785110950469971, + 0.6994727253913879 + ], + "mean": [ + 2.577927505176127, + -2.1525618086425604, + 0.7003541659101092 + ], + "std": [ + 1.6720365948940972, + 1.3768414767482617, + 0.0009102921174316669 + ], + "q01": [ + -0.7605057497562686, + -5.518671947805291, + 0.6999764703640262 + ], + "q99": [ + 6.038879433964306, + -0.5168299981147954, + 0.7029182461269848 + ] + }, + "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, + 0.24782371371751144, + 0.5971093380149053 + ], + "std": [ + 0.0, + 0.0, + 0.6642559499983499, + 0.3751891785184018 + ], + "q01": [ + 0.0, + 0.0, + -1.0, + 4.194193576416142e-06 + ], + "q99": [ + 0.0, + 0.0, + 1.0, + 1.0 + ] + }, + "end_effector_position_relative": { + "max": [ + 0.8848381638526917, + 0.8569459915161133, + 1.053884744644165 + ], + "min": [ + -0.5828549861907959, + -0.8486236333847046, + -0.3446122705936432 + ], + "mean": [ + 0.25006971579930243, + -0.005716176032514759, + 0.4426027721228362 + ], + "std": [ + 0.15838384751435522, + 0.25141504139800513, + 0.20646816234620438 + ], + "q01": [ + -0.017003721077380903, + -0.47738320165276543, + 0.1917563409753314 + ], + "q99": [ + 0.5401029199137674, + 0.4706646609268599, + 0.6965586537529516 + ] + }, + "end_effector_rotation_relative": { + "max": [ + 0.9999949932098389, + 0.9997354745864868, + 0.9900747537612915, + 0.9735639095306396 + ], + "min": [ + -0.9999914169311523, + -0.9999884366989136, + -0.9975369572639465, + 6.754255821306288e-08 + ], + "mean": [ + -0.22531628862537245, + 0.045669797538445966, + -0.06098055586474368, + 0.12292005493282424 + ], + "std": [ + 0.8506341437448487, + 0.2880763358322842, + 0.30816573781739026, + 0.16366165154486362 + ], + "q01": [ + -0.9957510593732032, + -0.5584741016891311, + -0.5371647381159943, + 0.0013784815838340492 + ], + "q99": [ + 0.9899375209218226, + 0.6327878884078038, + 0.5071539024531994, + 0.43555975763914345 + ] + }, + "gripper_qpos": { + "max": [ + 0.05289580300450325, + 0.00937709677964449 + ], + "min": [ + -0.011669117026031017, + -0.06307756155729294 + ], + "mean": [ + 0.03196506153371015, + -0.03229791582989581 + ], + "std": [ + 0.01171640503534049, + 0.01140016755425199 + ], + "q01": [ + 0.006290779996034005, + -0.04059717954551848 + ], + "q99": [ + 0.04038165838477291, + -0.006363076424932449 + ] + } + }, + "action": { + "base_motion": { + "max": [ + 1.0, + 1.0, + 1.0, + 0.0 + ], + "min": [ + -1.0, + -1.0, + -1.0, + 0.0 + ], + "mean": [ + 0.006112990930475949, + 0.002504955933349057, + -0.001639409958610569, + 0.0 + ], + "std": [ + 0.11478077265783658, + 0.1002662368589144, + 0.0807258815139021, + 0.0 + ], + "q01": [ + -0.16969970921613756, + -0.19919537649977875, + -0.15012139866371593, + 0.0 + ], + "q99": [ + 0.2466625453958827, + 0.25040842772679267, + 0.12777701094549537, + 0.0 + ] + }, + "control_mode": { + "max": [ + 1.0 + ], + "min": [ + -1.0 + ], + "mean": [ + -0.9253067479599785 + ], + "std": [ + 0.3792397014797737 + ], + "q01": [ + -1.0 + ], + "q99": [ + -0.1974093068624371 + ] + }, + "end_effector_position": { + "max": [ + 1.0, + 1.0, + 1.0 + ], + "min": [ + -1.0, + -1.0, + -1.0 + ], + "mean": [ + -0.006754222051360652, + 0.0027854853184013867, + -0.08087572130330477 + ], + "std": [ + 0.45779905420789546, + 0.40578975219522756, + 0.4062694098132682 + ], + "q01": [ + -0.968295960290036, + -0.9240402337368848, + -0.8249359802202183 + ], + "q99": [ + 0.8046634915851651, + 0.9212243025654695, + 0.838433711437593 + ] + }, + "end_effector_rotation": { + "max": [ + 1.0, + 1.0, + 1.0 + ], + "min": [ + -1.0, + -1.0, + -1.0 + ], + "mean": [ + 0.008633749251857347, + -0.020406457524911718, + -0.0027639446362883006 + ], + "std": [ + 0.11393972872407104, + 0.13300186268009506, + 0.12192288302920105 + ], + "q01": [ + -0.2668579353776187, + -0.40299251206771697, + -0.34712957892172586 + ], + "q99": [ + 0.32748281264383927, + 0.3179028969292779, + 0.33459405510046736 + ] + }, + "gripper_close": { + "max": [ + 1.0 + ], + "min": [ + -1.0 + ], + "mean": [ + -0.3091635769453286 + ], + "std": [ + 0.9510122432876397 + ], + "q01": [ + -1.0 + ], + "q99": [ + 0.7643545615216344 + ] + } + } + }, + "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