diff --git a/FewShot/39tasks/5shot/FromRKD-Timewarp_FullactionHead/seed0/default/train_5shot_seed0_fullactionhead_5k/checkpoint-5000/config.json b/FewShot/39tasks/5shot/FromRKD-Timewarp_FullactionHead/seed0/default/train_5shot_seed0_fullactionhead_5k/checkpoint-5000/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-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-Timewarp_FullactionHead/seed0/default/train_5shot_seed0_fullactionhead_5k/checkpoint-5000/experiment_cfg/metadata.json b/FewShot/39tasks/5shot/FromRKD-Timewarp_FullactionHead/seed0/default/train_5shot_seed0_fullactionhead_5k/checkpoint-5000/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-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-Timewarp_FullactionHead/seed0/default/train_5shot_seed0_fullactionhead_5k/checkpoint-5000/model-00001-of-00002.safetensors b/FewShot/39tasks/5shot/FromRKD-Timewarp_FullactionHead/seed0/default/train_5shot_seed0_fullactionhead_5k/checkpoint-5000/model-00001-of-00002.safetensors new file mode 100644 index 0000000000000000000000000000000000000000..1ad4ef9c9871007be569e79cbe2f56819e07abfc --- /dev/null +++ b/FewShot/39tasks/5shot/FromRKD-Timewarp_FullactionHead/seed0/default/train_5shot_seed0_fullactionhead_5k/checkpoint-5000/model-00001-of-00002.safetensors @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:dcbecde0056a2bcabba205866d2569d78defe9a1c298b5a74bc2dd650788b7f8 +size 4999367032 diff --git a/FewShot/39tasks/5shot/FromRKD-Timewarp_FullactionHead/seed0/default/train_5shot_seed0_fullactionhead_5k/checkpoint-5000/model-00002-of-00002.safetensors b/FewShot/39tasks/5shot/FromRKD-Timewarp_FullactionHead/seed0/default/train_5shot_seed0_fullactionhead_5k/checkpoint-5000/model-00002-of-00002.safetensors new file mode 100644 index 0000000000000000000000000000000000000000..e91e474591e7ac00c58886c3716958b3d36476ac --- /dev/null +++ b/FewShot/39tasks/5shot/FromRKD-Timewarp_FullactionHead/seed0/default/train_5shot_seed0_fullactionhead_5k/checkpoint-5000/model-00002-of-00002.safetensors @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:b335aecb764efe0190c9a584c730cc341c0b92034c62e6d34aaaca37abaa0435 +size 2664540348 diff --git a/FewShot/39tasks/5shot/FromRKD-Timewarp_FullactionHead/seed0/default/train_5shot_seed0_fullactionhead_5k/checkpoint-5000/model.safetensors.index.json b/FewShot/39tasks/5shot/FromRKD-Timewarp_FullactionHead/seed0/default/train_5shot_seed0_fullactionhead_5k/checkpoint-5000/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-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-Timewarp_FullactionHead/seed0/default/train_5shot_seed0_fullactionhead_5k/checkpoint-5000/optimizer.pt b/FewShot/39tasks/5shot/FromRKD-Timewarp_FullactionHead/seed0/default/train_5shot_seed0_fullactionhead_5k/checkpoint-5000/optimizer.pt new file mode 100644 index 0000000000000000000000000000000000000000..71fbfb56a0ef6f815c98b515ca650cce56c698df --- /dev/null +++ b/FewShot/39tasks/5shot/FromRKD-Timewarp_FullactionHead/seed0/default/train_5shot_seed0_fullactionhead_5k/checkpoint-5000/optimizer.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:2c70c414c4729531fb412e583f2788ff8f89e272adf909ccd9542309b788eb86 +size 8563326263 diff --git a/FewShot/39tasks/5shot/FromRKD-Timewarp_FullactionHead/seed0/default/train_5shot_seed0_fullactionhead_5k/checkpoint-5000/rng_state.pth b/FewShot/39tasks/5shot/FromRKD-Timewarp_FullactionHead/seed0/default/train_5shot_seed0_fullactionhead_5k/checkpoint-5000/rng_state.pth new file mode 100644 index 0000000000000000000000000000000000000000..426f2a5fb4bb19ed865194d2fd96eccc76554b2e --- /dev/null +++ b/FewShot/39tasks/5shot/FromRKD-Timewarp_FullactionHead/seed0/default/train_5shot_seed0_fullactionhead_5k/checkpoint-5000/rng_state.pth @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:a8a05114ecb0bdd72a49b35902c9e4ae20e5a0a84102f27cff35c6a71fc9ed70 +size 14645 diff --git a/FewShot/39tasks/5shot/FromRKD-Timewarp_FullactionHead/seed0/default/train_5shot_seed0_fullactionhead_5k/checkpoint-5000/scheduler.pt b/FewShot/39tasks/5shot/FromRKD-Timewarp_FullactionHead/seed0/default/train_5shot_seed0_fullactionhead_5k/checkpoint-5000/scheduler.pt new file mode 100644 index 0000000000000000000000000000000000000000..8cb18c42a19c524a920dd1ec232118cfa1c2aa5a --- /dev/null +++ b/FewShot/39tasks/5shot/FromRKD-Timewarp_FullactionHead/seed0/default/train_5shot_seed0_fullactionhead_5k/checkpoint-5000/scheduler.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:39f93e93eb5653777630c590a20e98e9373f48f53da024592383b624fc2dcbde +size 1465 diff --git a/FewShot/39tasks/5shot/FromRKD-Timewarp_FullactionHead/seed0/default/train_5shot_seed0_fullactionhead_5k/checkpoint-5000/trainer_state.json b/FewShot/39tasks/5shot/FromRKD-Timewarp_FullactionHead/seed0/default/train_5shot_seed0_fullactionhead_5k/checkpoint-5000/trainer_state.json new file mode 100644 index 0000000000000000000000000000000000000000..de222f2da206583f46fc08dec29efff7be65ce96 --- /dev/null +++ b/FewShot/39tasks/5shot/FromRKD-Timewarp_FullactionHead/seed0/default/train_5shot_seed0_fullactionhead_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": 0.189423368871212, + "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 + }, + { + "epoch": 5.050335570469799, + "flow_matching_loss": 0.013273283420130611, + "grad_norm": 0.2285473346710205, + "learning_rate": 1.1231228239056467e-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": 0.0058656760025769476, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0025024123024195433, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.014233764167875052, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.013273283420130611, + "weighted_rkd_loss": 0.007116882083937526 + }, + { + "epoch": 5.067114093959732, + "flow_matching_loss": 0.01667082030326128, + "grad_norm": 0.18787281215190887, + "learning_rate": 1.113528559434845e-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": 0.0063343644607812164, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.002568350301589817, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.015237079281359912, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.01667082030326128, + "weighted_rkd_loss": 0.007618539640679956 + }, + { + "epoch": 5.083892617449664, + "flow_matching_loss": 0.016702202148735523, + "grad_norm": 0.15388451516628265, + "learning_rate": 1.1039512004768476e-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": 0.006060790619812906, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0028450632351450624, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01496664434671402, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.016702202148735523, + "weighted_rkd_loss": 0.00748332217335701 + }, + { + "epoch": 5.100671140939597, + "flow_matching_loss": 0.014557135012000798, + "grad_norm": 0.22521626949310303, + "learning_rate": 1.0943911659763612e-05, + "loss": 0.021, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.0053807373624295, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.002053827058989555, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.012815301772207022, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.014557135012000798, + "weighted_rkd_loss": 0.006407650886103511 + }, + { + "epoch": 5.117449664429531, + "flow_matching_loss": 0.014089683070778847, + "grad_norm": 0.170135036110878, + "learning_rate": 1.084848874120266e-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": 0.005050548096187412, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.002229406835976988, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.012330503016710282, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.014089683070778847, + "weighted_rkd_loss": 0.006165251508355141 + }, + { + "epoch": 5.134228187919463, + "flow_matching_loss": 0.016467312071472406, + "grad_norm": 0.21445611119270325, + "learning_rate": 1.0753247423193196e-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": 0.0052245419006794695, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0022483804379589854, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.012697464134544135, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.016467312071472406, + "weighted_rkd_loss": 0.006348732067272067 + }, + { + "epoch": 5.151006711409396, + "flow_matching_loss": 0.014306401833891869, + "grad_norm": 0.1555250585079193, + "learning_rate": 1.0658191871899012e-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": 0.005630516982637346, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.002733483538031578, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01399451745674014, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.014306401833891869, + "weighted_rkd_loss": 0.00699725872837007 + }, + { + "epoch": 5.167785234899329, + "flow_matching_loss": 0.015325135178864002, + "grad_norm": 0.1826591193675995, + "learning_rate": 1.0563326245357852e-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": 0.006399042368866503, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.002548683516215533, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.015346768125891686, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.015325135178864002, + "weighted_rkd_loss": 0.007673384062945843 + }, + { + "epoch": 5.184563758389261, + "flow_matching_loss": 0.014437953382730484, + "grad_norm": 0.22793908417224884, + "learning_rate": 1.0468654693299557e-05, + "loss": 0.0205, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.005095488857477903, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0019791597384028136, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.012170137371867895, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.014437953382730484, + "weighted_rkd_loss": 0.0060850686859339476 + }, + { + "epoch": 5.201342281879195, + "flow_matching_loss": 0.014770141616463662, + "grad_norm": 0.1612703651189804, + "learning_rate": 1.0374181356964501e-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": 0.00542408530600369, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.002520810440182686, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.013368981052190065, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.014770141616463662, + "weighted_rkd_loss": 0.0066844905260950325 + }, + { + "epoch": 5.218120805369128, + "flow_matching_loss": 0.013219647854566575, + "grad_norm": 0.17939068377017975, + "learning_rate": 1.0279910368922465e-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": 0.005109355645254254, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.002146319497842342, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.012365030962973833, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.013219647854566575, + "weighted_rkd_loss": 0.006182515481486917 + }, + { + "epoch": 5.23489932885906, + "flow_matching_loss": 0.014013351593166589, + "grad_norm": 0.14087234437465668, + "learning_rate": 1.018584585289186e-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": 0.0053634805837646125, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.002066045394167304, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.012793006561696529, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.014013351593166589, + "weighted_rkd_loss": 0.006396503280848264 + }, + { + "epoch": 5.251677852348993, + "flow_matching_loss": 0.01542151253670454, + "grad_norm": 0.17132516205310822, + "learning_rate": 1.0091991923559348e-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": 0.00560535320546478, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0023774179979227485, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.013588124234229327, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.01542151253670454, + "weighted_rkd_loss": 0.0067940621171146635 + }, + { + "epoch": 5.268456375838926, + "flow_matching_loss": 0.014316082559525966, + "grad_norm": 0.15824805200099945, + "learning_rate": 9.99835268639984e-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": 0.005268195387907326, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0021136697265319525, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.012650060374289751, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.014316082559525966, + "weighted_rkd_loss": 0.006325030187144876 + }, + { + "epoch": 5.285234899328859, + "flow_matching_loss": 0.013256329577416182, + "grad_norm": 0.19563530385494232, + "learning_rate": 9.904932237496913e-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": 0.004841019283048809, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.002148597198538482, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.011830635741353035, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.013256329577416182, + "weighted_rkd_loss": 0.005915317870676518 + }, + { + "epoch": 5.302013422818792, + "flow_matching_loss": 0.012772800587117671, + "grad_norm": 0.23978853225708008, + "learning_rate": 9.811734663363659e-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": 0.005200855177827179, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.002329256245866418, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.012730966554954648, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.012772800587117671, + "weighted_rkd_loss": 0.006365483277477324 + }, + { + "epoch": 5.318791946308725, + "flow_matching_loss": 0.014975246787071229, + "grad_norm": 0.15939439833164215, + "learning_rate": 9.718764040763875e-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": 0.00513989080209285, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0018280613352544607, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.012107842974364758, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.014975246787071229, + "weighted_rkd_loss": 0.006053921487182379 + }, + { + "epoch": 5.3355704697986575, + "flow_matching_loss": 0.013651922205463051, + "grad_norm": 0.1433558464050293, + "learning_rate": 9.626024436533777e-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": 0.005451225582510233, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.002152232511434704, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.013054683711379767, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.013651922205463051, + "weighted_rkd_loss": 0.006527341855689883 + }, + { + "epoch": 5.35234899328859, + "flow_matching_loss": 0.01535201519727707, + "grad_norm": 0.19105803966522217, + "learning_rate": 9.533519907404095e-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": 0.004941799910739064, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0020117807027418165, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.011895380541682243, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.01535201519727707, + "weighted_rkd_loss": 0.0059476902708411215 + }, + { + "epoch": 5.369127516778524, + "flow_matching_loss": 0.015476669743657112, + "grad_norm": 0.1691632866859436, + "learning_rate": 9.4412544998226e-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": 0.004367403616197407, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.002023308025673032, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.010758115258067847, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.015476669743657112, + "weighted_rkd_loss": 0.0053790576290339235 + }, + { + "epoch": 5.385906040268456, + "flow_matching_loss": 0.015114418836310506, + "grad_norm": 0.15927359461784363, + "learning_rate": 9.349232249777124e-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": 0.005120966560207307, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.001963970740325749, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.012205903977155685, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.015114418836310506, + "weighted_rkd_loss": 0.0061029519885778425 + }, + { + "epoch": 5.402684563758389, + "flow_matching_loss": 0.01784304054453969, + "grad_norm": 0.24193723499774933, + "learning_rate": 9.257457182618985e-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": 0.004995469911955297, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0023999007069505753, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.012390840519219636, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.01784304054453969, + "weighted_rkd_loss": 0.006195420259609818 + }, + { + "epoch": 5.419463087248322, + "flow_matching_loss": 0.013066306710243225, + "grad_norm": 0.16006220877170563, + "learning_rate": 9.165933312886944e-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": 0.004924082010984421, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.002115910965949297, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.011964075081050397, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.013066306710243225, + "weighted_rkd_loss": 0.005982037540525198 + }, + { + "epoch": 5.436241610738255, + "flow_matching_loss": 0.014801208395510912, + "grad_norm": 0.16490410268306732, + "learning_rate": 9.07466464413155e-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": 0.005378607613965869, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0020040612551383676, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.012761276494711638, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.014801208395510912, + "weighted_rkd_loss": 0.006380638247355819 + }, + { + "epoch": 5.453020134228188, + "flow_matching_loss": 0.014771063253283501, + "grad_norm": 0.25417137145996094, + "learning_rate": 8.983655168740048e-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": 0.005886884219944477, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0022363172494806348, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.014010085724294186, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.014771063253283501, + "weighted_rkd_loss": 0.007005042862147093 + }, + { + "epoch": 5.469798657718121, + "flow_matching_loss": 0.013944575842469931, + "grad_norm": 0.15955711901187897, + "learning_rate": 8.892908867761718e-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": 0.006365548493340612, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0026807011221535505, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.0154117981903255, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.013944575842469931, + "weighted_rkd_loss": 0.00770589909516275 + }, + { + "epoch": 5.4865771812080535, + "flow_matching_loss": 0.01557155018672347, + "grad_norm": 0.148858442902565, + "learning_rate": 8.802429710733745e-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": 0.004635971668176353, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0020682581351138653, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.011340201552957296, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.01557155018672347, + "weighted_rkd_loss": 0.005670100776478648 + }, + { + "epoch": 5.503355704697986, + "flow_matching_loss": 0.01386803649365902, + "grad_norm": 0.15989872813224792, + "learning_rate": 8.71222165550757e-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": 0.004535936331376433, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0017340362304821611, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.010805909056216478, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.01386803649365902, + "weighted_rkd_loss": 0.005402954528108239 + }, + { + "epoch": 5.52013422818792, + "flow_matching_loss": 0.013872923143208028, + "grad_norm": 0.20790772140026093, + "learning_rate": 8.622288648075748e-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": 0.005408620904199779, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0021506744320504366, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.012967916205525398, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.013872923143208028, + "weighted_rkd_loss": 0.006483958102762699 + }, + { + "epoch": 5.5369127516778525, + "flow_matching_loss": 0.015501438919454814, + "grad_norm": 0.15546298027038574, + "learning_rate": 8.532634622399391e-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": 0.004575241915881633, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0018669651937671006, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.011017449013888837, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.015501438919454814, + "weighted_rkd_loss": 0.005508724506944418 + }, + { + "epoch": 5.553691275167785, + "flow_matching_loss": 0.012488964200019836, + "grad_norm": 0.23867888748645782, + "learning_rate": 8.44326350023601e-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": 0.005004889238625765, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0023841552902013063, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.012393934046849608, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.012488964200019836, + "weighted_rkd_loss": 0.006196967023424804 + }, + { + "epoch": 5.570469798657718, + "flow_matching_loss": 0.013480441831052303, + "grad_norm": 0.1565876305103302, + "learning_rate": 8.354179190968002e-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": 0.004363967292010784, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.001519673701841384, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.010247608181089163, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.013480441831052303, + "weighted_rkd_loss": 0.005123804090544581 + }, + { + "epoch": 5.587248322147651, + "flow_matching_loss": 0.015796675253659485, + "grad_norm": 0.24322588741779327, + "learning_rate": 8.265385591431661e-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": 0.005845426116138697, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0023771880776621403, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01406804034486413, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.015796675253659485, + "weighted_rkd_loss": 0.007034020172432065 + }, + { + "epoch": 5.604026845637584, + "flow_matching_loss": 0.01234276294708252, + "grad_norm": 0.16083206236362457, + "learning_rate": 8.176886585746671e-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": 0.004430905706249178, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0018418193329125642, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.010703630931675433, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.01234276294708252, + "weighted_rkd_loss": 0.005351815465837717 + }, + { + "epoch": 5.620805369127517, + "flow_matching_loss": 0.01186593621969223, + "grad_norm": 0.20607060194015503, + "learning_rate": 8.088686045146247e-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": 0.004728447226807475, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.001986333087552339, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.011443227529525757, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.01186593621969223, + "weighted_rkd_loss": 0.005721613764762878 + }, + { + "epoch": 5.6375838926174495, + "flow_matching_loss": 0.013174599409103394, + "grad_norm": 0.1966661512851715, + "learning_rate": 8.00078782780775e-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": 0.004724739631637931, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.001865789678413421, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.011315269116312265, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.013174599409103394, + "weighted_rkd_loss": 0.005657634558156132 + }, + { + "epoch": 5.654362416107382, + "flow_matching_loss": 0.01585042094811797, + "grad_norm": 0.245307058095932, + "learning_rate": 7.913195778683979e-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": 0.005080026038922369, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.002161143010016531, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.012321195006370545, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.01585042094811797, + "weighted_rkd_loss": 0.006160597503185272 + }, + { + "epoch": 5.671140939597316, + "flow_matching_loss": 0.015605000220239163, + "grad_norm": 0.20907120406627655, + "learning_rate": 7.825913729334908e-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": 0.005415506870485842, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0021290503675118087, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.012960064318031072, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.015605000220239163, + "weighted_rkd_loss": 0.006480032159015536 + }, + { + "epoch": 5.6879194630872485, + "flow_matching_loss": 0.016725842375308275, + "grad_norm": 0.14447364211082458, + "learning_rate": 7.73894549776013e-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": 0.00547388365957886, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0023031249875202775, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.013250892329961062, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.016725842375308275, + "weighted_rkd_loss": 0.006625446164980531 + }, + { + "epoch": 5.704697986577181, + "flow_matching_loss": 0.01270838575437665, + "grad_norm": 0.23588861525058746, + "learning_rate": 7.65229488823184e-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": 0.005150177539326251, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0022166476352140307, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.012517002783715725, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.01270838575437665, + "weighted_rkd_loss": 0.006258501391857863 + }, + { + "epoch": 5.721476510067114, + "flow_matching_loss": 0.015314796008169651, + "grad_norm": 0.182273268699646, + "learning_rate": 7.565965691128396e-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": 0.00467446381226182, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0020193661213852463, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01136829387396574, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.015314796008169651, + "weighted_rkd_loss": 0.00568414693698287 + }, + { + "epoch": 5.7382550335570475, + "flow_matching_loss": 0.013604261912405491, + "grad_norm": 0.23861733078956604, + "learning_rate": 7.47996168276856e-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": 0.004217569250613451, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0016527582542039455, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.010087896650657058, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.013604261912405491, + "weighted_rkd_loss": 0.005043948325328529 + }, + { + "epoch": 5.75503355704698, + "flow_matching_loss": 0.014357882179319858, + "grad_norm": 0.1535225212574005, + "learning_rate": 7.394286625246258e-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": 0.0049197893822565675, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0016938321408815683, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.011533410940319299, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.014357882179319858, + "weighted_rkd_loss": 0.0057667054701596495 + }, + { + "epoch": 5.771812080536913, + "flow_matching_loss": 0.0134419372305274, + "grad_norm": 0.15331624448299408, + "learning_rate": 7.308944266266063e-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": 0.004597988841123879, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0018919462454505265, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01108792396262288, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.0134419372305274, + "weighted_rkd_loss": 0.00554396198131144 + }, + { + "epoch": 5.7885906040268456, + "flow_matching_loss": 0.014776341710239648, + "grad_norm": 0.25563931465148926, + "learning_rate": 7.223938338979233e-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": 0.004740951280109584, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0018884244142100214, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.011370326951146125, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.014776341710239648, + "weighted_rkd_loss": 0.005685163475573063 + }, + { + "epoch": 5.805369127516778, + "flow_matching_loss": 0.014994413778185844, + "grad_norm": 0.15913286805152893, + "learning_rate": 7.139272561820397e-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": 0.004294173908419907, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0015233114478178321, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.010111659159883856, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.014994413778185844, + "weighted_rkd_loss": 0.005055829579941928 + }, + { + "epoch": 5.822147651006711, + "flow_matching_loss": 0.012550541758537292, + "grad_norm": 0.171083003282547, + "learning_rate": 7.054950638344936e-06, + "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": 0.004584924993105232, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0019210978411138059, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.011090947780758142, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.012550541758537292, + "weighted_rkd_loss": 0.005545473890379071 + }, + { + "epoch": 5.8389261744966445, + "flow_matching_loss": 0.01349291019141674, + "grad_norm": 0.17222721874713898, + "learning_rate": 6.970976257066941e-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": 0.005571219907142222, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.00232730865245685, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.013469748198986053, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.01349291019141674, + "weighted_rkd_loss": 0.006734874099493027 + }, + { + "epoch": 5.855704697986577, + "flow_matching_loss": 0.013362313434481621, + "grad_norm": 0.26061469316482544, + "learning_rate": 6.887353091297896e-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": 0.005095311347395182, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0020249326713383197, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.012215555505827069, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.013362313434481621, + "weighted_rkd_loss": 0.0061077777529135345 + }, + { + "epoch": 5.87248322147651, + "flow_matching_loss": 0.014484726265072823, + "grad_norm": 0.16677652299404144, + "learning_rate": 6.804084798985965e-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": 0.0047961494885385035, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.001770413643680513, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.011362712737172842, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.014484726265072823, + "weighted_rkd_loss": 0.005681356368586421 + }, + { + "epoch": 5.889261744966443, + "flow_matching_loss": 0.015598151553422212, + "grad_norm": 0.2078307569026947, + "learning_rate": 6.721175022556009e-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": 0.005224377172999084, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0020208677859045567, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01246962221339345, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.015598151553422212, + "weighted_rkd_loss": 0.006234811106696725 + }, + { + "epoch": 5.906040268456376, + "flow_matching_loss": 0.01374701363965869, + "grad_norm": 0.2645686864852905, + "learning_rate": 6.638627388750245e-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": 0.0053967045387253165, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0019286571186967194, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.012722066324204206, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.01374701363965869, + "weighted_rkd_loss": 0.006361033162102103 + }, + { + "epoch": 5.922818791946309, + "flow_matching_loss": 0.016289659217000006, + "grad_norm": 0.2069074511528015, + "learning_rate": 6.556445508469583e-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": 0.0047795751597732306, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0018127361487131565, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.011371886404231191, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.016289659217000006, + "weighted_rkd_loss": 0.005685943202115596 + }, + { + "epoch": 5.939597315436242, + "flow_matching_loss": 0.014452755730599166, + "grad_norm": 0.1726139336824417, + "learning_rate": 6.474632976615708e-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": 0.0047163198003545405, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0022801060462370516, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.011712745716795325, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.014452755730599166, + "weighted_rkd_loss": 0.005856372858397662 + }, + { + "epoch": 5.956375838926174, + "flow_matching_loss": 0.013426850829273463, + "grad_norm": 0.1568567007780075, + "learning_rate": 6.393193371933791e-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": 0.005156427063047886, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0024364317068830132, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.012749285716563464, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.013426850829273463, + "weighted_rkd_loss": 0.006374642858281732 + }, + { + "epoch": 5.973154362416107, + "flow_matching_loss": 0.013179479539394379, + "grad_norm": 0.1591392308473587, + "learning_rate": 6.312130256855977e-06, + "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": 0.0041168633382767435, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0016634542611427606, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.009897180832922459, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.013179479539394379, + "weighted_rkd_loss": 0.004948590416461229 + }, + { + "epoch": 5.989932885906041, + "flow_matching_loss": 0.01301817810162902, + "grad_norm": 0.12511791288852692, + "learning_rate": 6.23144717734552e-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": 0.004697659728117287, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0018042239942587912, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.011199543299153448, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.01301817810162902, + "weighted_rkd_loss": 0.005599771649576724 + }, + { + "epoch": 6.006711409395973, + "flow_matching_loss": 0.01299028513021767, + "grad_norm": 0.1949714571237564, + "learning_rate": 6.151147662741695e-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": 0.004801549506373704, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.002034022466978058, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.011637121578678489, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.01299028513021767, + "weighted_rkd_loss": 0.0058185607893392445 + }, + { + "epoch": 6.023489932885906, + "flow_matching_loss": 0.014248680043965577, + "grad_norm": 0.18050147593021393, + "learning_rate": 6.071235225605415e-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": 0.004431381216272711, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0018716220394708217, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.010734384413808584, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.014248680043965577, + "weighted_rkd_loss": 0.005367192206904292 + }, + { + "epoch": 6.040268456375839, + "flow_matching_loss": 0.014908248744904994, + "grad_norm": 0.14633774757385254, + "learning_rate": 5.991713361565546e-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": 0.004304599575698375, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0017973661771975458, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.010406565340235829, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.014908248744904994, + "weighted_rkd_loss": 0.005203282670117914 + }, + { + "epoch": 6.057046979865772, + "flow_matching_loss": 0.012182641122490168, + "grad_norm": 0.231786847114563, + "learning_rate": 5.912585549166051e-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": 0.0053547847550362345, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0023295242921449245, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.013039093744009734, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.012182641122490168, + "weighted_rkd_loss": 0.006519546872004867 + }, + { + "epoch": 6.073825503355705, + "flow_matching_loss": 0.01349649541079998, + "grad_norm": 0.15987418591976166, + "learning_rate": 5.833855249713777e-06, + "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": 0.0037824744707904757, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0015456155641004443, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.009110564505681396, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.01349649541079998, + "weighted_rkd_loss": 0.004555282252840698 + }, + { + "epoch": 6.090604026845638, + "flow_matching_loss": 0.014953711070120335, + "grad_norm": 0.17302507162094116, + "learning_rate": 5.7555259071270645e-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": 0.004880516487173736, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.001975899317767471, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.011736932350322605, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.014953711070120335, + "weighted_rkd_loss": 0.005868466175161302 + }, + { + "epoch": 6.10738255033557, + "flow_matching_loss": 0.012672061333432794, + "grad_norm": 0.1585019826889038, + "learning_rate": 5.677600947785117e-06, + "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": 0.0042997630778700115, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0017347801825962962, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.010334306443110109, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.012672061333432794, + "weighted_rkd_loss": 0.0051671532215550545 + }, + { + "epoch": 6.124161073825503, + "flow_matching_loss": 0.013576794881373644, + "grad_norm": 0.1438910812139511, + "learning_rate": 5.600083780378099e-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": 0.0039490525843575595, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.001519838348031044, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.009417943563312292, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.013576794881373644, + "weighted_rkd_loss": 0.004708971781656146 + }, + { + "epoch": 6.140939597315437, + "flow_matching_loss": 0.013494531717151403, + "grad_norm": 0.19352352619171143, + "learning_rate": 5.522977795758038e-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": 0.004718074877746403, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0016840157913975418, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.011120165511965752, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.013494531717151403, + "weighted_rkd_loss": 0.005560082755982876 + }, + { + "epoch": 6.157718120805369, + "flow_matching_loss": 0.014294523280113935, + "grad_norm": 0.1627226024866104, + "learning_rate": 5.4462863667904805e-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": 0.00501574594527483, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0020766448462381957, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.012108136806637049, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.014294523280113935, + "weighted_rkd_loss": 0.0060540684033185245 + }, + { + "epoch": 6.174496644295302, + "flow_matching_loss": 0.015006358921527862, + "grad_norm": 0.18388888239860535, + "learning_rate": 5.370012848206983e-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": 0.005103649152442813, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.002023331017699093, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.012230629660189151, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.015006358921527862, + "weighted_rkd_loss": 0.006115314830094576 + }, + { + "epoch": 6.191275167785235, + "flow_matching_loss": 0.01274308622814715, + "grad_norm": 0.19488470256328583, + "learning_rate": 5.29416057645834e-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": 0.004091479768976569, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0017668359796516597, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.00994979557581246, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.01274308622814715, + "weighted_rkd_loss": 0.00497489778790623 + }, + { + "epoch": 6.208053691275167, + "flow_matching_loss": 0.013387705152854323, + "grad_norm": 0.1380261331796646, + "learning_rate": 5.218732869568635e-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": 0.004827187396585941, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0020433113211765887, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.011697686091065407, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.013387705152854323, + "weighted_rkd_loss": 0.005848843045532704 + }, + { + "epoch": 6.224832214765101, + "flow_matching_loss": 0.011581274913623929, + "grad_norm": 0.15014977753162384, + "learning_rate": 5.14373302699013e-06, + "loss": 0.0167, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.004214526782743633, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0017126499151345342, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.010141703393310308, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.011581274913623929, + "weighted_rkd_loss": 0.005070851696655154 + }, + { + "epoch": 6.241610738255034, + "flow_matching_loss": 0.013908269163221122, + "grad_norm": 0.1939009577035904, + "learning_rate": 5.069164329458894e-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": 0.0045234568417072294, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0020099034416489305, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.011056817043572665, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.013908269163221122, + "weighted_rkd_loss": 0.005528408521786332 + }, + { + "epoch": 6.258389261744966, + "flow_matching_loss": 0.014544932637363672, + "grad_norm": 0.20686961710453033, + "learning_rate": 4.995030038851342e-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": 0.0044759223237633705, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0015815795748494565, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.010533424327149987, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.014544932637363672, + "weighted_rkd_loss": 0.005266712163574994 + }, + { + "epoch": 6.275167785234899, + "flow_matching_loss": 0.012062178738415242, + "grad_norm": 0.141107439994812, + "learning_rate": 4.9213333980415e-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": 0.0045945595251396295, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0018156210309825838, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.011004740232601761, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.012062178738415242, + "weighted_rkd_loss": 0.005502370116300881 + }, + { + "epoch": 6.291946308724833, + "flow_matching_loss": 0.013018049858510494, + "grad_norm": 0.147357776761055, + "learning_rate": 4.848077630759175e-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": 0.004304882767610252, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0018924566567875445, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.010502222133800388, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.013018049858510494, + "weighted_rkd_loss": 0.005251111066900194 + }, + { + "epoch": 6.308724832214765, + "flow_matching_loss": 0.014812441915273667, + "grad_norm": 0.21773049235343933, + "learning_rate": 4.775265941448951e-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": 0.004822571738623082, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0017652085982263089, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.011410352075472474, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.014812441915273667, + "weighted_rkd_loss": 0.005705176037736237 + }, + { + "epoch": 6.325503355704698, + "flow_matching_loss": 0.013156043365597725, + "grad_norm": 0.15065298974514008, + "learning_rate": 4.702901515129989e-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": 0.004030594695359468, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0016236922470852732, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.009684881567955017, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.013156043365597725, + "weighted_rkd_loss": 0.004842440783977509 + }, + { + "epoch": 6.342281879194631, + "flow_matching_loss": 0.01194716515019536, + "grad_norm": 0.18975025415420532, + "learning_rate": 4.630987517256734e-06, + "loss": 0.0164, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.0037176207872107623, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.001498365867882967, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.008933607395738364, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.01194716515019536, + "weighted_rkd_loss": 0.004466803697869182 + }, + { + "epoch": 6.359060402684563, + "flow_matching_loss": 0.013762779394164681, + "grad_norm": 0.17298924922943115, + "learning_rate": 4.559527093580412e-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": 0.004396662302315235, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0016519619966857135, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.010445286706089973, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.013762779394164681, + "weighted_rkd_loss": 0.005222643353044986 + }, + { + "epoch": 6.375838926174497, + "flow_matching_loss": 0.013622605893760919, + "grad_norm": 0.13369767367839813, + "learning_rate": 4.488523370011456e-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": 0.003639714792370796, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0014581867260858416, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.008737616240978241, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.013622605893760919, + "weighted_rkd_loss": 0.0043688081204891205 + }, + { + "epoch": 6.39261744966443, + "flow_matching_loss": 0.012390201818197966, + "grad_norm": 0.13894836604595184, + "learning_rate": 4.417979452482766e-06, + "loss": 0.017, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.003824461461044848, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.001623190485406667, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.00927211344242096, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.012390201818197966, + "weighted_rkd_loss": 0.00463605672121048 + }, + { + "epoch": 6.409395973154362, + "flow_matching_loss": 0.011436420166864992, + "grad_norm": 0.13240008056163788, + "learning_rate": 4.347898426813811e-06, + "loss": 0.0172, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.004742332524619997, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.002084954851306975, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01156962001696229, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.011436420166864992, + "weighted_rkd_loss": 0.005784810008481145 + }, + { + "epoch": 6.426174496644295, + "flow_matching_loss": 0.012280825292691588, + "grad_norm": 0.14240750670433044, + "learning_rate": 4.278283358575694e-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": 0.00448585560079664, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0018520746380090714, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.010823785979300737, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.012280825292691588, + "weighted_rkd_loss": 0.005411892989650368 + }, + { + "epoch": 6.442953020134228, + "flow_matching_loss": 0.012999581918120385, + "grad_norm": 0.14199872314929962, + "learning_rate": 4.2091372929570145e-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": 0.004581688530743122, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0020241894177161156, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.011187566444277763, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.012999581918120385, + "weighted_rkd_loss": 0.005593783222138882 + }, + { + "epoch": 6.459731543624161, + "flow_matching_loss": 0.015608004108071328, + "grad_norm": 0.17531105875968933, + "learning_rate": 4.1404632546306934e-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": 0.004717425419948995, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0018084937881212681, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01124334461055696, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.015608004108071328, + "weighted_rkd_loss": 0.00562167230527848 + }, + { + "epoch": 6.476510067114094, + "flow_matching_loss": 0.01236470346339047, + "grad_norm": 0.19072523713111877, + "learning_rate": 4.0722642476216286e-06, + "loss": 0.0171, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.003976859035901725, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.001551550440490246, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.009505268465727568, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.01236470346339047, + "weighted_rkd_loss": 0.004752634232863784 + }, + { + "epoch": 6.493288590604027, + "flow_matching_loss": 0.013742083543911576, + "grad_norm": 0.13357122242450714, + "learning_rate": 4.00454325517533e-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": 0.004746155813336372, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0017761539667844773, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.011268465546891093, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.013742083543911576, + "weighted_rkd_loss": 0.005634232773445547 + }, + { + "epoch": 6.510067114093959, + "flow_matching_loss": 0.01476223086938262, + "grad_norm": 0.2286960482597351, + "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": 0.004158663819544018, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0018114029197022318, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.010128730582073331, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.01476223086938262, + "weighted_rkd_loss": 0.005064365291036666 + }, + { + "epoch": 6.526845637583893, + "flow_matching_loss": 0.013768768729642033, + "grad_norm": 0.36107489466667175, + "learning_rate": 3.870547142273918e-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": 0.005147815262898803, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0018881296273320913, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.012183760106563569, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.013768768729642033, + "weighted_rkd_loss": 0.006091880053281784 + }, + { + "epoch": 6.543624161073826, + "flow_matching_loss": 0.013367434125393629, + "grad_norm": 0.2914619743824005, + "learning_rate": 3.8042778832428736e-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": 0.004395952657796442, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0020264136604964734, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.010818319115787744, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.013367434125393629, + "weighted_rkd_loss": 0.005409159557893872 + }, + { + "epoch": 6.560402684563758, + "flow_matching_loss": 0.017430961970239876, + "grad_norm": 0.159344881772995, + "learning_rate": 3.7384983613663195e-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": 0.0042855771724134685, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0016469669179059565, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.010218121297657489, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.017430961970239876, + "weighted_rkd_loss": 0.0051090606488287445 + }, + { + "epoch": 6.577181208053691, + "flow_matching_loss": 0.014220637455582618, + "grad_norm": 0.1419205665588379, + "learning_rate": 3.6732114540536503e-06, + "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": 0.003935556556098163, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.001527376810554415, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.009398489771410824, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.014220637455582618, + "weighted_rkd_loss": 0.004699244885705412 + }, + { + "epoch": 6.593959731543624, + "flow_matching_loss": 0.013942512311041355, + "grad_norm": 0.18272189795970917, + "learning_rate": 3.608420017165683e-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": 0.0038367455126717687, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.001534297107718885, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.0092077880166471, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.013942512311041355, + "weighted_rkd_loss": 0.00460389400832355 + }, + { + "epoch": 6.610738255033557, + "flow_matching_loss": 0.015249136649072171, + "grad_norm": 0.16941304504871368, + "learning_rate": 3.5441268848897664e-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": 0.00395207954570651, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0015867861919105053, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.009490945329889654, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.015249136649072171, + "weighted_rkd_loss": 0.004745472664944827 + }, + { + "epoch": 6.62751677852349, + "flow_matching_loss": 0.012341193947941064, + "grad_norm": 0.1718492954969406, + "learning_rate": 3.480334869615792e-06, + "loss": 0.0173, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.004190959827974438, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0015963940415531397, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.009978313744068146, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.012341193947941064, + "weighted_rkd_loss": 0.004989156872034073 + }, + { + "epoch": 6.644295302013423, + "flow_matching_loss": 0.013152943830937147, + "grad_norm": 0.16254186630249023, + "learning_rate": 3.417046761813155e-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": 0.004420910519547761, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0017289884388446807, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.010570809524506331, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.013152943830937147, + "weighted_rkd_loss": 0.005285404762253166 + }, + { + "epoch": 6.6610738255033555, + "flow_matching_loss": 0.01244001667946577, + "grad_norm": 0.13000580668449402, + "learning_rate": 3.3542653299087263e-06, + "loss": 0.0173, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.004057411407120526, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0015844591427594424, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.009699281957000493, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.01244001667946577, + "weighted_rkd_loss": 0.004849640978500247 + }, + { + "epoch": 6.677852348993289, + "flow_matching_loss": 0.010493615129962563, + "grad_norm": 0.15268608927726746, + "learning_rate": 3.2919933201657117e-06, + "loss": 0.0159, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.004557304014451802, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.001655701722484082, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01077030966989696, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.010493615129962563, + "weighted_rkd_loss": 0.00538515483494848 + }, + { + "epoch": 6.694630872483222, + "flow_matching_loss": 0.013269064854830504, + "grad_norm": 0.18566147983074188, + "learning_rate": 3.2302334565635584e-06, + "loss": 0.0179, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.003922327677719295, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0013941960758529603, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.009238851442933082, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.013269064854830504, + "weighted_rkd_loss": 0.004619425721466541 + }, + { + "epoch": 6.7114093959731544, + "flow_matching_loss": 0.012347688805311919, + "grad_norm": 0.1489165872335434, + "learning_rate": 3.1689884406787674e-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": 0.004499934636987746, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0015349635505117476, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.010534832859411836, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.012347688805311919, + "weighted_rkd_loss": 0.005267416429705918 + }, + { + "epoch": 6.728187919463087, + "flow_matching_loss": 0.012257662788033485, + "grad_norm": 0.1501564085483551, + "learning_rate": 3.1082609515667424e-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": 0.005181431001983583, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.001985070831142366, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.012347932904958725, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.012257662788033485, + "weighted_rkd_loss": 0.0061739664524793625 + }, + { + "epoch": 6.74496644295302, + "flow_matching_loss": 0.014351728092879056, + "grad_norm": 0.17460209131240845, + "learning_rate": 3.048053645644594e-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": 0.004749392857775092, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0019455971429124474, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.011444382835179567, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.014351728092879056, + "weighted_rkd_loss": 0.005722191417589784 + }, + { + "epoch": 6.7617449664429525, + "flow_matching_loss": 0.012328914739191532, + "grad_norm": 0.18522049486637115, + "learning_rate": 2.988369156574916e-06, + "loss": 0.0169, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.0039043401600793006, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0013507586263585836, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.00915943905711174, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.012328914739191532, + "weighted_rkd_loss": 0.00457971952855587 + }, + { + "epoch": 6.778523489932886, + "flow_matching_loss": 0.012781216204166413, + "grad_norm": 0.15217529237270355, + "learning_rate": 2.9292100951506225e-06, + "loss": 0.0168, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.003358302265405655, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0013781144050881266, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.008094718912616372, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.012781216204166413, + "weighted_rkd_loss": 0.004047359456308186 + }, + { + "epoch": 6.795302013422819, + "flow_matching_loss": 0.012864326685667038, + "grad_norm": 0.19025535881519318, + "learning_rate": 2.8705790491807033e-06, + "loss": 0.0173, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.003772235964424908, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.001426016999175772, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.008970488933846355, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.012864326685667038, + "weighted_rkd_loss": 0.004485244466923177 + }, + { + "epoch": 6.8120805369127515, + "flow_matching_loss": 0.012307855812832714, + "grad_norm": 0.19388285279273987, + "learning_rate": 2.8124785833770433e-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": 0.004534709919244051, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.001737466419581324, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.010806886246427894, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.012307855812832714, + "weighted_rkd_loss": 0.005403443123213947 + }, + { + "epoch": 6.828859060402684, + "flow_matching_loss": 0.015203884430229664, + "grad_norm": 0.15002429485321045, + "learning_rate": 2.754911239242241e-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": 0.003685287549160421, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0016135861864313484, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.008984161168336868, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.015203884430229664, + "weighted_rkd_loss": 0.004492080584168434 + }, + { + "epoch": 6.845637583892618, + "flow_matching_loss": 0.012085525365546346, + "grad_norm": 0.11901740729808807, + "learning_rate": 2.6978795349584204e-06, + "loss": 0.0159, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.0031661784276366236, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0012743374507408589, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.007606694335117936, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.012085525365546346, + "weighted_rkd_loss": 0.003803347167558968 + }, + { + "epoch": 6.8624161073825505, + "flow_matching_loss": 0.01438704738393426, + "grad_norm": 0.16812318563461304, + "learning_rate": 2.641385965277088e-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": 0.003867070819251239, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0017184988828375936, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.009452640358358622, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.01438704738393426, + "weighted_rkd_loss": 0.004726320179179311 + }, + { + "epoch": 6.879194630872483, + "flow_matching_loss": 0.012696548271924258, + "grad_norm": 0.14939044415950775, + "learning_rate": 2.5854330014099923e-06, + "loss": 0.0171, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.0036327670561149715, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.001501097227446735, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.008766631409525872, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.012696548271924258, + "weighted_rkd_loss": 0.004383315704762936 + }, + { + "epoch": 6.895973154362416, + "flow_matching_loss": 0.013077494082972407, + "grad_norm": 0.17584958672523499, + "learning_rate": 2.5300230909210298e-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": 0.004372379346750677, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.001655038306489587, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01039979699999094, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.013077494082972407, + "weighted_rkd_loss": 0.00519989849999547 + }, + { + "epoch": 6.912751677852349, + "flow_matching_loss": 0.01232621893286705, + "grad_norm": 0.14323139190673828, + "learning_rate": 2.4751586576191983e-06, + "loss": 0.0168, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.003784377779811621, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0013999658403918147, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.008968721283599735, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.01232621893286705, + "weighted_rkd_loss": 0.004484360641799867 + }, + { + "epoch": 6.929530201342282, + "flow_matching_loss": 0.011209582397714257, + "grad_norm": 0.15581539273262024, + "learning_rate": 2.420842101452538e-06, + "loss": 0.0167, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.004541526525281369, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.001841145200887695, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.010924198292195797, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.011209582397714257, + "weighted_rkd_loss": 0.005462099146097898 + }, + { + "epoch": 6.946308724832215, + "flow_matching_loss": 0.011785591254010797, + "grad_norm": 0.16155187785625458, + "learning_rate": 2.3670757984031816e-06, + "loss": 0.016, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.0035794392693787812, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0013535585603676736, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.008512437157332898, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.011785591254010797, + "weighted_rkd_loss": 0.004256218578666449 + }, + { + "epoch": 6.9630872483221475, + "flow_matching_loss": 0.010752106131985784, + "grad_norm": 0.2019570916891098, + "learning_rate": 2.313862100383404e-06, + "loss": 0.0158, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.004264209792017937, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0015931188769172876, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.010121538490056991, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.010752106131985784, + "weighted_rkd_loss": 0.0050607692450284954 + }, + { + "epoch": 6.97986577181208, + "flow_matching_loss": 0.010426551662385463, + "grad_norm": 0.1295207440853119, + "learning_rate": 2.2612033351327406e-06, + "loss": 0.0148, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.003634925140067935, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0014228249609004706, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.008692675270140171, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.010426551662385463, + "weighted_rkd_loss": 0.0043463376350700855 + }, + { + "epoch": 6.996644295302014, + "flow_matching_loss": 0.012410965654999018, + "grad_norm": 0.15585634112358093, + "learning_rate": 2.2091018061161788e-06, + "loss": 0.0171, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.0038361755898222325, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0016191505710594356, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.009291501902043819, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.012410965654999018, + "weighted_rkd_loss": 0.004645750951021909 + }, + { + "epoch": 7.0134228187919465, + "flow_matching_loss": 0.013164480961859227, + "grad_norm": 0.16433636844158173, + "learning_rate": 2.15755979242338e-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": 0.003910991130396724, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0018357451888732612, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.009657727554440498, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.013164480961859227, + "weighted_rkd_loss": 0.004828863777220249 + }, + { + "epoch": 7.030201342281879, + "flow_matching_loss": 0.013598700612783432, + "grad_norm": 0.16424773633480072, + "learning_rate": 2.106579548669e-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": 0.004430685797706247, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0016169998794794082, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.010478371428325773, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.013598700612783432, + "weighted_rkd_loss": 0.005239185714162886 + }, + { + "epoch": 7.046979865771812, + "flow_matching_loss": 0.011333533329889178, + "grad_norm": 0.15079383552074432, + "learning_rate": 2.0561633048940477e-06, + "loss": 0.0161, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.003885891393292695, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0017129390267655253, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.009484721766784786, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.011333533329889178, + "weighted_rkd_loss": 0.004742360883392393 + }, + { + "epoch": 7.063758389261745, + "flow_matching_loss": 0.013496446702629328, + "grad_norm": 0.14999589323997498, + "learning_rate": 2.006313266468355e-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": 0.004368492448702454, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0016905601136386395, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.010427544871345163, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.013496446702629328, + "weighted_rkd_loss": 0.0052137724356725815 + }, + { + "epoch": 7.080536912751678, + "flow_matching_loss": 0.012888733576983214, + "grad_norm": 0.18489472568035126, + "learning_rate": 1.957031613994097e-06, + "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": 0.0038338485173881054, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.001534275064477697, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.00920197213999927, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.012888733576983214, + "weighted_rkd_loss": 0.004600986069999635 + }, + { + "epoch": 7.097315436241611, + "flow_matching_loss": 0.012950423266738653, + "grad_norm": 0.13276706635951996, + "learning_rate": 1.9083205032103995e-06, + "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": 0.0038131119683384894, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0015088815591298044, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.009135105554014445, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.012950423266738653, + "weighted_rkd_loss": 0.004567552777007222 + }, + { + "epoch": 7.114093959731544, + "flow_matching_loss": 0.013198824971914292, + "grad_norm": 0.1334235966205597, + "learning_rate": 1.8601820648990536e-06, + "loss": 0.0174, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.003466808726079762, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0014216269366443158, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.008355244342237712, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.013198824971914292, + "weighted_rkd_loss": 0.004177622171118856 + }, + { + "epoch": 7.130872483221476, + "flow_matching_loss": 0.012214665487408637, + "grad_norm": 0.17523948848247528, + "learning_rate": 1.8126184047912936e-06, + "loss": 0.0163, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.003384548705071211, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0013439524045679718, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.008113049762323498, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.012214665487408637, + "weighted_rkd_loss": 0.004056524881161749 + }, + { + "epoch": 7.14765100671141, + "flow_matching_loss": 0.011672427924349905, + "grad_norm": 0.12189946323633194, + "learning_rate": 1.7656316034757024e-06, + "loss": 0.0162, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.0036959612276405095, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0017431621032301336, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.00913508445955813, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.011672427924349905, + "weighted_rkd_loss": 0.004567542229779065 + }, + { + "epoch": 7.1644295302013425, + "flow_matching_loss": 0.012925780657678843, + "grad_norm": 0.17717908322811127, + "learning_rate": 1.7192237163071816e-06, + "loss": 0.0171, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.003533428453374654, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0013367140141781419, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.008403570856899023, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.012925780657678843, + "weighted_rkd_loss": 0.0042017854284495115 + }, + { + "epoch": 7.181208053691275, + "flow_matching_loss": 0.013103020470589399, + "grad_norm": 0.11885635554790497, + "learning_rate": 1.6733967733170573e-06, + "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": 0.003977324068546295, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0015103903540875762, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.00946503859013319, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.013103020470589399, + "weighted_rkd_loss": 0.004732519295066595 + }, + { + "epoch": 7.197986577181208, + "flow_matching_loss": 0.010397190414369106, + "grad_norm": 0.16860656440258026, + "learning_rate": 1.628152779124279e-06, + "loss": 0.0151, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.0038975286413915457, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0016264894686173648, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.009421546664088964, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.010397190414369106, + "weighted_rkd_loss": 0.004710773332044482 + }, + { + "epoch": 7.214765100671141, + "flow_matching_loss": 0.012783728912472725, + "grad_norm": 0.12303238362073898, + "learning_rate": 1.583493712847715e-06, + "loss": 0.0174, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.003879416105337441, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0015254201600328088, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.009284252347424627, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.012783728912472725, + "weighted_rkd_loss": 0.0046421261737123135 + }, + { + "epoch": 7.231543624161074, + "flow_matching_loss": 0.013865695055574178, + "grad_norm": 0.1526777744293213, + "learning_rate": 1.5394215280196084e-06, + "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": 0.003514062752947211, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0013878538622520864, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.00841597942635417, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.013865695055574178, + "weighted_rkd_loss": 0.004207989713177085 + }, + { + "epoch": 7.248322147651007, + "flow_matching_loss": 0.013047731202095746, + "grad_norm": 0.16690459847450256, + "learning_rate": 1.4959381525000904e-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": 0.004131821263581514, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0015723746037110686, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.009836017154157162, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.013047731202095746, + "weighted_rkd_loss": 0.004918008577078581 + }, + { + "epoch": 7.26510067114094, + "flow_matching_loss": 0.013816094864159823, + "grad_norm": 0.1267736554145813, + "learning_rate": 1.453045488392879e-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": 0.003801518469117582, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0012555589317344129, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.008858596067875623, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.013816094864159823, + "weighted_rkd_loss": 0.004429298033937812 + }, + { + "epoch": 7.281879194630872, + "flow_matching_loss": 0.012940576393157243, + "grad_norm": 0.13604986667633057, + "learning_rate": 1.4107454119620484e-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": 0.004049134370870888, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0014924665214493872, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.009590735286474228, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.012940576393157243, + "weighted_rkd_loss": 0.004795367643237114 + }, + { + "epoch": 7.298657718120805, + "flow_matching_loss": 0.013075377326458693, + "grad_norm": 0.18737582862377167, + "learning_rate": 1.3690397735499778e-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": 0.004112353594973683, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0015578099992126226, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.009782517235726118, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.013075377326458693, + "weighted_rkd_loss": 0.004891258617863059 + }, + { + "epoch": 7.315436241610739, + "flow_matching_loss": 0.013750377995893358, + "grad_norm": 0.19063593447208405, + "learning_rate": 1.327930397496403e-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": 0.003749578760471195, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0014163357904180885, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.008915493311360479, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.013750377995893358, + "weighted_rkd_loss": 0.0044577466556802395 + }, + { + "epoch": 7.332214765100671, + "flow_matching_loss": 0.010600294126197696, + "grad_norm": 0.12065929174423218, + "learning_rate": 1.287419082058604e-06, + "loss": 0.0151, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.00381252421066165, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.001365220220759511, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.00899026868864894, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.010600294126197696, + "weighted_rkd_loss": 0.00449513434432447 + }, + { + "epoch": 7.348993288590604, + "flow_matching_loss": 0.011868054792284966, + "grad_norm": 0.13562294840812683, + "learning_rate": 1.2475075993327589e-06, + "loss": 0.0162, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.003603435377590358, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0015282828360795974, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.008735153544694185, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.011868054792284966, + "weighted_rkd_loss": 0.004367576772347092 + }, + { + "epoch": 7.365771812080537, + "flow_matching_loss": 0.011605020193383098, + "grad_norm": 0.1373264491558075, + "learning_rate": 1.208197695176413e-06, + "loss": 0.0162, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.0038801249815151096, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0015233642479870468, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.009283614344894886, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.011605020193383098, + "weighted_rkd_loss": 0.004641807172447443 + }, + { + "epoch": 7.382550335570469, + "flow_matching_loss": 0.01095195272937417, + "grad_norm": 0.1840600073337555, + "learning_rate": 1.1694910891321259e-06, + "loss": 0.0158, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.004077260848134756, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0014926740492228418, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.009647195786237716, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.01095195272937417, + "weighted_rkd_loss": 0.004823597893118858 + }, + { + "epoch": 7.399328859060403, + "flow_matching_loss": 0.012686336040496826, + "grad_norm": 0.24466432631015778, + "learning_rate": 1.1313894743522363e-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": 0.004122675489634275, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0017858694423921405, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.010031220503151417, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.012686336040496826, + "weighted_rkd_loss": 0.005015610251575708 + }, + { + "epoch": 7.416107382550336, + "flow_matching_loss": 0.012575617153197527, + "grad_norm": 0.17857523262500763, + "learning_rate": 1.0938945175248016e-06, + "loss": 0.0169, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.003530106134712696, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0015555102261714637, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.008615722553804517, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.012575617153197527, + "weighted_rkd_loss": 0.004307861276902259 + }, + { + "epoch": 7.432885906040268, + "flow_matching_loss": 0.014236471243202686, + "grad_norm": 0.11501364409923553, + "learning_rate": 1.0570078588007086e-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": 0.0043713633203879, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0016129569092299789, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.010355683555826545, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.014236471243202686, + "weighted_rkd_loss": 0.005177841777913273 + }, + { + "epoch": 7.449664429530201, + "flow_matching_loss": 0.01142531242221594, + "grad_norm": 0.19348150491714478, + "learning_rate": 1.0207311117219009e-06, + "loss": 0.0167, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.00433938808273524, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0017794008366763593, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.010458177095279098, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.01142531242221594, + "weighted_rkd_loss": 0.005229088547639549 + }, + { + "epoch": 7.466442953020135, + "flow_matching_loss": 0.011046114470809698, + "grad_norm": 0.17002864181995392, + "learning_rate": 9.850658631508198e-07, + "loss": 0.0155, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.003723083692602813, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0014892871375195682, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.008935454487800597, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.011046114470809698, + "weighted_rkd_loss": 0.004467727243900299 + }, + { + "epoch": 7.483221476510067, + "flow_matching_loss": 0.012161614559590816, + "grad_norm": 0.15816764533519745, + "learning_rate": 9.500136732009751e-07, + "loss": 0.0172, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.004277206980623305, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0014976980164647102, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.010052111838012934, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.012161614559590816, + "weighted_rkd_loss": 0.005026055919006467 + }, + { + "epoch": 7.5, + "flow_matching_loss": 0.011525012785568833, + "grad_norm": 0.1495368480682373, + "learning_rate": 9.155760751687054e-07, + "loss": 0.0158, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.0035916498396545648, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0013772728387266397, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.00856057251803577, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.011525012785568833, + "weighted_rkd_loss": 0.004280286259017885 + }, + { + "epoch": 7.516778523489933, + "flow_matching_loss": 0.011997128324583172, + "grad_norm": 0.1800948828458786, + "learning_rate": 8.817545754661155e-07, + "loss": 0.017, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.004208199260756373, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0015775962849147618, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.009993994794785976, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.011997128324583172, + "weighted_rkd_loss": 0.004996997397392988 + }, + { + "epoch": 7.533557046979865, + "flow_matching_loss": 0.011707661580294371, + "grad_norm": 0.18044085800647736, + "learning_rate": 8.485506535551668e-07, + "loss": 0.0163, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.003826503362506628, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0015013229451142252, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.009154329635202884, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.011707661580294371, + "weighted_rkd_loss": 0.004577164817601442 + }, + { + "epoch": 7.550335570469799, + "flow_matching_loss": 0.011094614583998918, + "grad_norm": 0.1431879699230194, + "learning_rate": 8.159657618829685e-07, + "loss": 0.0154, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.0035614509135484694, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0014412890886887908, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.008564190845936537, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.011094614583998918, + "weighted_rkd_loss": 0.004282095422968269 + }, + { + "epoch": 7.567114093959732, + "flow_matching_loss": 0.01384103624150157, + "grad_norm": 0.1595686376094818, + "learning_rate": 7.840013258182471e-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": 0.003950062114745379, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0015118620882276445, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.009411986311897635, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.01384103624150157, + "weighted_rkd_loss": 0.0047059931559488176 + }, + { + "epoch": 7.583892617449664, + "flow_matching_loss": 0.01329559963196516, + "grad_norm": 0.17983092367649078, + "learning_rate": 7.526587435889809e-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": 0.004146448709070683, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.001896186772501096, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.010189084196463227, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.01329559963196516, + "weighted_rkd_loss": 0.0050945420982316135 + }, + { + "epoch": 7.600671140939597, + "flow_matching_loss": 0.012888224283233284, + "grad_norm": 0.1517045646905899, + "learning_rate": 7.219393862212593e-07, + "loss": 0.0174, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.0037810103734955192, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0014952955476474016, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.0090573163703084, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.012888224283233284, + "weighted_rkd_loss": 0.0045286581851542 + }, + { + "epoch": 7.617449664429531, + "flow_matching_loss": 0.01240743980742991, + "grad_norm": 0.19681622087955475, + "learning_rate": 6.918445974792819e-07, + "loss": 0.0173, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.004042636440135539, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0017130049236584455, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.009798277774825693, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.01240743980742991, + "weighted_rkd_loss": 0.004899138887412846 + }, + { + "epoch": 7.634228187919463, + "flow_matching_loss": 0.012322596600279212, + "grad_norm": 0.1485264003276825, + "learning_rate": 6.6237569380661e-07, + "loss": 0.0167, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.003632089029997587, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0015278115111868828, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.00879198950715363, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.012322596600279212, + "weighted_rkd_loss": 0.004395994753576815 + }, + { + "epoch": 7.651006711409396, + "flow_matching_loss": 0.012507512792944908, + "grad_norm": 0.21264411509037018, + "learning_rate": 6.335339642685495e-07, + "loss": 0.0174, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.004084891243837774, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.001712805643910542, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.009882588125765323, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.012507512792944908, + "weighted_rkd_loss": 0.0049412940628826615 + }, + { + "epoch": 7.667785234899329, + "flow_matching_loss": 0.011168187530711294, + "grad_norm": 0.16030879318714142, + "learning_rate": 6.05320670495787e-07, + "loss": 0.0154, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.0035492916591465472, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0014431263203732668, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.008541709743440151, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.011168187530711294, + "weighted_rkd_loss": 0.004270854871720076 + }, + { + "epoch": 7.684563758389261, + "flow_matching_loss": 0.012330411188304424, + "grad_norm": 0.13624970614910126, + "learning_rate": 5.777370466291954e-07, + "loss": 0.0168, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.003753046621568501, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0015292260563001038, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.00903531932272017, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.012330411188304424, + "weighted_rkd_loss": 0.004517659661360085 + }, + { + "epoch": 7.701342281879195, + "flow_matching_loss": 0.011879453249275685, + "grad_norm": 0.14977416396141052, + "learning_rate": 5.507842992658336e-07, + "loss": 0.0155, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.0029674590798094867, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0013971363950986415, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.007332054572179914, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.011879453249275685, + "weighted_rkd_loss": 0.003666027286089957 + }, + { + "epoch": 7.718120805369128, + "flow_matching_loss": 0.013211170118302107, + "grad_norm": 0.153207927942276, + "learning_rate": 5.24463607406192e-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": 0.003935256344266236, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0013785569346509874, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.00924906968139112, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.013211170118302107, + "weighted_rkd_loss": 0.00462453484069556 + }, + { + "epoch": 7.73489932885906, + "flow_matching_loss": 0.011284657474607229, + "grad_norm": 0.1558067351579666, + "learning_rate": 4.987761224025944e-07, + "loss": 0.0163, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.004252262297086418, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.001456828135997057, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.00996135277673602, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.011284657474607229, + "weighted_rkd_loss": 0.00498067638836801 + }, + { + "epoch": 7.751677852348993, + "flow_matching_loss": 0.015760020352900028, + "grad_norm": 0.15146303176879883, + "learning_rate": 4.737229679088556e-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": 0.0038813820108771326, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0013940333214122803, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.009156797314062715, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.015760020352900028, + "weighted_rkd_loss": 0.004578398657031357 + }, + { + "epoch": 7.768456375838926, + "flow_matching_loss": 0.01323346383869648, + "grad_norm": 0.18961799144744873, + "learning_rate": 4.493052398311115e-07, + "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": 0.0036372275091707705, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.001513058808632195, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.008787513757124544, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.01323346383869648, + "weighted_rkd_loss": 0.004393756878562272 + }, + { + "epoch": 7.785234899328859, + "flow_matching_loss": 0.012718648929148913, + "grad_norm": 0.13715103268623352, + "learning_rate": 4.2552400627989043e-07, + "loss": 0.0172, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.003695989469997585, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0016332714585587382, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.009025250328704715, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.012718648929148913, + "weighted_rkd_loss": 0.004512625164352357 + }, + { + "epoch": 7.802013422818792, + "flow_matching_loss": 0.013017370831221342, + "grad_norm": 0.15020433068275452, + "learning_rate": 4.023803075233923e-07, + "loss": 0.0169, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.0032229928066954015, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0014095393009483815, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.007855524914339185, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.013017370831221342, + "weighted_rkd_loss": 0.0039277624571695926 + }, + { + "epoch": 7.818791946308725, + "flow_matching_loss": 0.013933033216744661, + "grad_norm": 0.11856479197740555, + "learning_rate": 3.7987515594196997e-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": 0.004223480657674372, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.001563948963303119, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.010010910406708717, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.013933033216744661, + "weighted_rkd_loss": 0.005005455203354358 + }, + { + "epoch": 7.8355704697986575, + "flow_matching_loss": 0.01209373315796256, + "grad_norm": 0.16292829811573029, + "learning_rate": 3.580095359838648e-07, + "loss": 0.0162, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.0034446867881342768, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0013271984236780554, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.008216572040691972, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.01209373315796256, + "weighted_rkd_loss": 0.004108286020345986 + }, + { + "epoch": 7.85234899328859, + "flow_matching_loss": 0.011257783556357027, + "grad_norm": 0.12844575941562653, + "learning_rate": 3.3678440412212576e-07, + "loss": 0.0151, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.0032084846636280416, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0011872378352563828, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.007604207145050168, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.011257783556357027, + "weighted_rkd_loss": 0.003802103572525084 + }, + { + "epoch": 7.869127516778524, + "flow_matching_loss": 0.013048629881814123, + "grad_norm": 0.1467456817626953, + "learning_rate": 3.162006888127833e-07, + "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": 0.004155136458575725, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0016613823710940779, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.009971655253320933, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.013048629881814123, + "weighted_rkd_loss": 0.0049858276266604665 + }, + { + "epoch": 7.885906040268456, + "flow_matching_loss": 0.014346472080796957, + "grad_norm": 0.12790009379386902, + "learning_rate": 2.9625929045422327e-07, + "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": 0.003967941668815911, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0014618574175983667, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.009397740941494703, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.014346472080796957, + "weighted_rkd_loss": 0.004698870470747352 + }, + { + "epoch": 7.902684563758389, + "flow_matching_loss": 0.01271637286990881, + "grad_norm": 0.16283373534679413, + "learning_rate": 2.7696108134781186e-07, + "loss": 0.0174, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.0039179531624540685, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0014930704259313644, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.009328976878896356, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.01271637286990881, + "weighted_rkd_loss": 0.004664488439448178 + }, + { + "epoch": 7.919463087248322, + "flow_matching_loss": 0.014059433806687593, + "grad_norm": 0.17524674534797668, + "learning_rate": 2.583069056597376e-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": 0.003773231990635395, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.001535862369928509, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.009082326386123895, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.014059433806687593, + "weighted_rkd_loss": 0.0045411631930619475 + }, + { + "epoch": 7.936241610738255, + "flow_matching_loss": 0.011765244835987686, + "grad_norm": 0.15201984345912933, + "learning_rate": 2.4029757938407116e-07, + "loss": 0.017, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.004287443915382027, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0018286535050719976, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.010403541289269925, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.011765244835987686, + "weighted_rkd_loss": 0.005201770644634962 + }, + { + "epoch": 7.953020134228188, + "flow_matching_loss": 0.013755477033555508, + "grad_norm": 0.19749130308628082, + "learning_rate": 2.229338903070871e-07, + "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": 0.0035599785624071956, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0014380234526470304, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.008557980647310615, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.013755477033555508, + "weighted_rkd_loss": 0.004278990323655307 + }, + { + "epoch": 7.969798657718121, + "flow_matching_loss": 0.014288169704377652, + "grad_norm": 0.17997831106185913, + "learning_rate": 2.0621659797279645e-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": 0.004082369967363775, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0016643157927319408, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.009829055797308684, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.014288169704377652, + "weighted_rkd_loss": 0.004914527898654342 + }, + { + "epoch": 7.9865771812080535, + "flow_matching_loss": 0.012982644187286497, + "grad_norm": 0.13272035121917725, + "learning_rate": 1.901464336497233e-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": 0.004022148088552058, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0015711100655607878, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.009615406207740308, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.012982644187286497, + "weighted_rkd_loss": 0.004807703103870154 + }, + { + "epoch": 8.003355704697986, + "flow_matching_loss": 0.012017120141535997, + "grad_norm": 0.1425294727087021, + "learning_rate": 1.7472410029891527e-07, + "loss": 0.0164, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.003690410265699029, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0013321076985448599, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.00871292818337679, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.012017120141535997, + "weighted_rkd_loss": 0.004356464091688395 + }, + { + "epoch": 8.020134228187919, + "flow_matching_loss": 0.011508079571649432, + "grad_norm": 0.1185975894331932, + "learning_rate": 1.5995027254319173e-07, + "loss": 0.0163, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.00405461045447737, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0014693920616991817, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.009578613005578518, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.011508079571649432, + "weighted_rkd_loss": 0.004789306502789259 + }, + { + "epoch": 8.036912751677852, + "flow_matching_loss": 0.011774006625637412, + "grad_norm": 0.12193300575017929, + "learning_rate": 1.4582559663764372e-07, + "loss": 0.0157, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.0032693485030904413, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0013653967354912311, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.00790409385226667, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.011774006625637412, + "weighted_rkd_loss": 0.003952046926133335 + }, + { + "epoch": 8.053691275167786, + "flow_matching_loss": 0.01376564335078001, + "grad_norm": 0.15403912961483002, + "learning_rate": 1.323506904413485e-07, + "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": 0.0037099570967257024, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0013822478882502764, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.008802162064239383, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.01376564335078001, + "weighted_rkd_loss": 0.004401081032119692 + }, + { + "epoch": 8.070469798657719, + "flow_matching_loss": 0.012571563199162483, + "grad_norm": 0.14847010374069214, + "learning_rate": 1.195261433903594e-07, + "loss": 0.0169, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.0035822264617308974, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0014063604292459786, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.008570813480764628, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.012571563199162483, + "weighted_rkd_loss": 0.004285406740382314 + }, + { + "epoch": 8.087248322147651, + "flow_matching_loss": 0.01303102783858776, + "grad_norm": 0.22479461133480072, + "learning_rate": 1.0735251647190969e-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": 0.004655194259248674, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0017578598228283227, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.011068248329684139, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.01303102783858776, + "weighted_rkd_loss": 0.005534124164842069 + }, + { + "epoch": 8.104026845637584, + "flow_matching_loss": 0.010825584968551993, + "grad_norm": 0.14044439792633057, + "learning_rate": 9.583034219987408e-08, + "loss": 0.0148, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.0032617375254631042, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0014044177252799273, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.007927892776206136, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.010825584968551993, + "weighted_rkd_loss": 0.003963946388103068 + }, + { + "epoch": 8.120805369127517, + "flow_matching_loss": 0.010878020897507668, + "grad_norm": 0.16086778044700623, + "learning_rate": 8.496012459148717e-08, + "loss": 0.0155, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.0038354229647666214, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0015930189460050316, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.009263864858075977, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.010878020897507668, + "weighted_rkd_loss": 0.004631932429037988 + }, + { + "epoch": 8.13758389261745, + "flow_matching_loss": 0.01375404316931963, + "grad_norm": 0.1751631796360016, + "learning_rate": 7.474233914527618e-08, + "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": 0.0038205950520932676, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0016478243051096797, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.00928901438601315, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.01375404316931963, + "weighted_rkd_loss": 0.004644507193006575 + }, + { + "epoch": 8.154362416107382, + "flow_matching_loss": 0.013494851160794496, + "grad_norm": 0.19929954409599304, + "learning_rate": 6.517743282028254e-08, + "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": 0.0034150957595556974, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0013123318669386208, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.008142523420974612, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.013494851160794496, + "weighted_rkd_loss": 0.004071261710487306 + }, + { + "epoch": 8.171140939597315, + "flow_matching_loss": 0.014098628470674158, + "grad_norm": 0.15721264481544495, + "learning_rate": 5.6265824016492604e-08, + "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": 0.0037533565191552043, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0015949857071973384, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.009101698687300087, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.014098628470674158, + "weighted_rkd_loss": 0.004550849343650043 + }, + { + "epoch": 8.187919463087248, + "flow_matching_loss": 0.013258166052401066, + "grad_norm": 0.12723150849342346, + "learning_rate": 4.800790255654719e-08, + "loss": 0.0173, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.0034274237463250757, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.001281585212564096, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.008136432617902756, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.013258166052401066, + "weighted_rkd_loss": 0.004068216308951378 + }, + { + "epoch": 8.204697986577182, + "flow_matching_loss": 0.01369173233397305, + "grad_norm": 0.21861012279987335, + "learning_rate": 4.040402966867862e-08, + "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": 0.0037710790522396563, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.001406640448840335, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.008948798477649688, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.01369173233397305, + "weighted_rkd_loss": 0.004474399238824844 + }, + { + "epoch": 8.221476510067115, + "flow_matching_loss": 0.011236160341650248, + "grad_norm": 0.1503625065088272, + "learning_rate": 3.3454537970925014e-08, + "loss": 0.0154, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.003518209815956652, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0013535513891838491, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.008389971125870943, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.011236160341650248, + "weighted_rkd_loss": 0.004194985562935472 + }, + { + "epoch": 8.238255033557047, + "flow_matching_loss": 0.013852568995207547, + "grad_norm": 0.16968147456645966, + "learning_rate": 2.715973145656192e-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": 0.0037199769169092177, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0014850097242742777, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.0089249636977911, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.013852568995207547, + "weighted_rkd_loss": 0.00446248184889555 + }, + { + "epoch": 8.25503355704698, + "flow_matching_loss": 0.011306985560804606, + "grad_norm": 0.14712770283222198, + "learning_rate": 2.151988548082129e-08, + "loss": 0.0158, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.003772911708801985, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0014175664517097175, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.008963389880955219, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.011306985560804606, + "weighted_rkd_loss": 0.0044816949404776095 + }, + { + "epoch": 8.271812080536913, + "flow_matching_loss": 0.012866842467337847, + "grad_norm": 0.14772823452949524, + "learning_rate": 1.6535246748837772e-08, + "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": 0.004458117461763322, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0017672449350357056, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.010683479905128478, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.012866842467337847, + "weighted_rkd_loss": 0.005341739952564239 + }, + { + "epoch": 8.288590604026846, + "flow_matching_loss": 0.013925638142973184, + "grad_norm": 0.17435146868228912, + "learning_rate": 1.2206033304855702e-08, + "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": 0.004066174430772662, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0015223037218675018, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.009654652560129762, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.013925638142973184, + "weighted_rkd_loss": 0.004827326280064881 + }, + { + "epoch": 8.305369127516778, + "flow_matching_loss": 0.013670520065352321, + "grad_norm": 0.22648420929908752, + "learning_rate": 8.532434522695031e-09, + "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": 0.004508587578311562, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0018952674232423305, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.010912442533299326, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.013670520065352321, + "weighted_rkd_loss": 0.005456221266649663 + }, + { + "epoch": 8.322147651006711, + "flow_matching_loss": 0.012735329382121564, + "grad_norm": 0.15040388703346252, + "learning_rate": 5.5146110974679805e-09, + "loss": 0.017, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.003590844781138003, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0014258236333262175, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.008607513178139925, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.012735329382121564, + "weighted_rkd_loss": 0.004303756589069963 + }, + { + "epoch": 8.338926174496644, + "flow_matching_loss": 0.011451990390196443, + "grad_norm": 0.1485985368490219, + "learning_rate": 3.1526950385496467e-09, + "loss": 0.0163, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.004111752426251769, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0014832181157544255, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.0097067229449749, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.011451990390196443, + "weighted_rkd_loss": 0.00485336147248745 + }, + { + "epoch": 8.355704697986576, + "flow_matching_loss": 0.011133158579468727, + "grad_norm": 0.15696369111537933, + "learning_rate": 1.4467896637959666e-09, + "loss": 0.0153, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.0035388516378588976, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0012879677873570473, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.008365671196952463, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.011133158579468727, + "weighted_rkd_loss": 0.0041828355984762314 + }, + { + "epoch": 8.37248322147651, + "flow_matching_loss": 0.013223122153431178, + "grad_norm": 0.17017419636249542, + "learning_rate": 3.9696959503898733e-10, + "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": 0.0038438115268945696, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.001413160125957802, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.00910078315064311, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.013223122153431178, + "weighted_rkd_loss": 0.004550391575321555 + }, + { + "epoch": 8.389261744966444, + "flow_matching_loss": 0.012224378576502205, + "grad_norm": 0.16432853043079376, + "learning_rate": 3.280754806156416e-12, + "loss": 0.0164, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.0034646759275346996, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0013615088420920074, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.008290860708802939, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.012224378576502205, + "weighted_rkd_loss": 0.004145430354401469 + } + ], + "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/train_5shot_seed0_fullactionhead_5k/config.json b/FewShot/39tasks/5shot/FromRKD-Timewarp_FullactionHead/seed0/default/train_5shot_seed0_fullactionhead_5k/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/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/model-00001-of-00002.safetensors b/FewShot/39tasks/5shot/FromRKD-Timewarp_FullactionHead/seed0/default/train_5shot_seed0_fullactionhead_5k/model-00001-of-00002.safetensors new file mode 100644 index 0000000000000000000000000000000000000000..1ad4ef9c9871007be569e79cbe2f56819e07abfc --- /dev/null +++ b/FewShot/39tasks/5shot/FromRKD-Timewarp_FullactionHead/seed0/default/train_5shot_seed0_fullactionhead_5k/model-00001-of-00002.safetensors @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:dcbecde0056a2bcabba205866d2569d78defe9a1c298b5a74bc2dd650788b7f8 +size 4999367032 diff --git a/FewShot/39tasks/5shot/FromRKD-Timewarp_FullactionHead/seed0/default/train_5shot_seed0_fullactionhead_5k/model-00002-of-00002.safetensors b/FewShot/39tasks/5shot/FromRKD-Timewarp_FullactionHead/seed0/default/train_5shot_seed0_fullactionhead_5k/model-00002-of-00002.safetensors new file mode 100644 index 0000000000000000000000000000000000000000..e91e474591e7ac00c58886c3716958b3d36476ac --- /dev/null +++ b/FewShot/39tasks/5shot/FromRKD-Timewarp_FullactionHead/seed0/default/train_5shot_seed0_fullactionhead_5k/model-00002-of-00002.safetensors @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:b335aecb764efe0190c9a584c730cc341c0b92034c62e6d34aaaca37abaa0435 +size 2664540348 diff --git a/FewShot/39tasks/5shot/FromRKD-Timewarp_FullactionHead/seed0/default/train_5shot_seed0_fullactionhead_5k/model.safetensors.index.json b/FewShot/39tasks/5shot/FromRKD-Timewarp_FullactionHead/seed0/default/train_5shot_seed0_fullactionhead_5k/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/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/trainer_state.json b/FewShot/39tasks/5shot/FromRKD-Timewarp_FullactionHead/seed0/default/train_5shot_seed0_fullactionhead_5k/trainer_state.json new file mode 100644 index 0000000000000000000000000000000000000000..d250fd89dc25b48453d3f754dd86de55af7b1eb6 --- /dev/null +++ b/FewShot/39tasks/5shot/FromRKD-Timewarp_FullactionHead/seed0/default/train_5shot_seed0_fullactionhead_5k/trainer_state.json @@ -0,0 +1,23043 @@ +{ + "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": 0.189423368871212, + "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 + }, + { + "epoch": 5.050335570469799, + "flow_matching_loss": 0.013273283420130611, + "grad_norm": 0.2285473346710205, + "learning_rate": 1.1231228239056467e-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": 0.0058656760025769476, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0025024123024195433, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.014233764167875052, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.013273283420130611, + "weighted_rkd_loss": 0.007116882083937526 + }, + { + "epoch": 5.067114093959732, + "flow_matching_loss": 0.01667082030326128, + "grad_norm": 0.18787281215190887, + "learning_rate": 1.113528559434845e-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": 0.0063343644607812164, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.002568350301589817, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.015237079281359912, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.01667082030326128, + "weighted_rkd_loss": 0.007618539640679956 + }, + { + "epoch": 5.083892617449664, + "flow_matching_loss": 0.016702202148735523, + "grad_norm": 0.15388451516628265, + "learning_rate": 1.1039512004768476e-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": 0.006060790619812906, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0028450632351450624, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01496664434671402, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.016702202148735523, + "weighted_rkd_loss": 0.00748332217335701 + }, + { + "epoch": 5.100671140939597, + "flow_matching_loss": 0.014557135012000798, + "grad_norm": 0.22521626949310303, + "learning_rate": 1.0943911659763612e-05, + "loss": 0.021, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.0053807373624295, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.002053827058989555, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.012815301772207022, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.014557135012000798, + "weighted_rkd_loss": 0.006407650886103511 + }, + { + "epoch": 5.117449664429531, + "flow_matching_loss": 0.014089683070778847, + "grad_norm": 0.170135036110878, + "learning_rate": 1.084848874120266e-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": 0.005050548096187412, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.002229406835976988, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.012330503016710282, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.014089683070778847, + "weighted_rkd_loss": 0.006165251508355141 + }, + { + "epoch": 5.134228187919463, + "flow_matching_loss": 0.016467312071472406, + "grad_norm": 0.21445611119270325, + "learning_rate": 1.0753247423193196e-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": 0.0052245419006794695, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0022483804379589854, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.012697464134544135, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.016467312071472406, + "weighted_rkd_loss": 0.006348732067272067 + }, + { + "epoch": 5.151006711409396, + "flow_matching_loss": 0.014306401833891869, + "grad_norm": 0.1555250585079193, + "learning_rate": 1.0658191871899012e-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": 0.005630516982637346, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.002733483538031578, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01399451745674014, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.014306401833891869, + "weighted_rkd_loss": 0.00699725872837007 + }, + { + "epoch": 5.167785234899329, + "flow_matching_loss": 0.015325135178864002, + "grad_norm": 0.1826591193675995, + "learning_rate": 1.0563326245357852e-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": 0.006399042368866503, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.002548683516215533, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.015346768125891686, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.015325135178864002, + "weighted_rkd_loss": 0.007673384062945843 + }, + { + "epoch": 5.184563758389261, + "flow_matching_loss": 0.014437953382730484, + "grad_norm": 0.22793908417224884, + "learning_rate": 1.0468654693299557e-05, + "loss": 0.0205, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.005095488857477903, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0019791597384028136, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.012170137371867895, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.014437953382730484, + "weighted_rkd_loss": 0.0060850686859339476 + }, + { + "epoch": 5.201342281879195, + "flow_matching_loss": 0.014770141616463662, + "grad_norm": 0.1612703651189804, + "learning_rate": 1.0374181356964501e-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": 0.00542408530600369, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.002520810440182686, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.013368981052190065, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.014770141616463662, + "weighted_rkd_loss": 0.0066844905260950325 + }, + { + "epoch": 5.218120805369128, + "flow_matching_loss": 0.013219647854566575, + "grad_norm": 0.17939068377017975, + "learning_rate": 1.0279910368922465e-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": 0.005109355645254254, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.002146319497842342, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.012365030962973833, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.013219647854566575, + "weighted_rkd_loss": 0.006182515481486917 + }, + { + "epoch": 5.23489932885906, + "flow_matching_loss": 0.014013351593166589, + "grad_norm": 0.14087234437465668, + "learning_rate": 1.018584585289186e-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": 0.0053634805837646125, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.002066045394167304, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.012793006561696529, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.014013351593166589, + "weighted_rkd_loss": 0.006396503280848264 + }, + { + "epoch": 5.251677852348993, + "flow_matching_loss": 0.01542151253670454, + "grad_norm": 0.17132516205310822, + "learning_rate": 1.0091991923559348e-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": 0.00560535320546478, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0023774179979227485, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.013588124234229327, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.01542151253670454, + "weighted_rkd_loss": 0.0067940621171146635 + }, + { + "epoch": 5.268456375838926, + "flow_matching_loss": 0.014316082559525966, + "grad_norm": 0.15824805200099945, + "learning_rate": 9.99835268639984e-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": 0.005268195387907326, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0021136697265319525, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.012650060374289751, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.014316082559525966, + "weighted_rkd_loss": 0.006325030187144876 + }, + { + "epoch": 5.285234899328859, + "flow_matching_loss": 0.013256329577416182, + "grad_norm": 0.19563530385494232, + "learning_rate": 9.904932237496913e-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": 0.004841019283048809, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.002148597198538482, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.011830635741353035, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.013256329577416182, + "weighted_rkd_loss": 0.005915317870676518 + }, + { + "epoch": 5.302013422818792, + "flow_matching_loss": 0.012772800587117671, + "grad_norm": 0.23978853225708008, + "learning_rate": 9.811734663363659e-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": 0.005200855177827179, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.002329256245866418, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.012730966554954648, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.012772800587117671, + "weighted_rkd_loss": 0.006365483277477324 + }, + { + "epoch": 5.318791946308725, + "flow_matching_loss": 0.014975246787071229, + "grad_norm": 0.15939439833164215, + "learning_rate": 9.718764040763875e-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": 0.00513989080209285, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0018280613352544607, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.012107842974364758, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.014975246787071229, + "weighted_rkd_loss": 0.006053921487182379 + }, + { + "epoch": 5.3355704697986575, + "flow_matching_loss": 0.013651922205463051, + "grad_norm": 0.1433558464050293, + "learning_rate": 9.626024436533777e-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": 0.005451225582510233, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.002152232511434704, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.013054683711379767, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.013651922205463051, + "weighted_rkd_loss": 0.006527341855689883 + }, + { + "epoch": 5.35234899328859, + "flow_matching_loss": 0.01535201519727707, + "grad_norm": 0.19105803966522217, + "learning_rate": 9.533519907404095e-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": 0.004941799910739064, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0020117807027418165, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.011895380541682243, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.01535201519727707, + "weighted_rkd_loss": 0.0059476902708411215 + }, + { + "epoch": 5.369127516778524, + "flow_matching_loss": 0.015476669743657112, + "grad_norm": 0.1691632866859436, + "learning_rate": 9.4412544998226e-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": 0.004367403616197407, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.002023308025673032, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.010758115258067847, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.015476669743657112, + "weighted_rkd_loss": 0.0053790576290339235 + }, + { + "epoch": 5.385906040268456, + "flow_matching_loss": 0.015114418836310506, + "grad_norm": 0.15927359461784363, + "learning_rate": 9.349232249777124e-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": 0.005120966560207307, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.001963970740325749, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.012205903977155685, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.015114418836310506, + "weighted_rkd_loss": 0.0061029519885778425 + }, + { + "epoch": 5.402684563758389, + "flow_matching_loss": 0.01784304054453969, + "grad_norm": 0.24193723499774933, + "learning_rate": 9.257457182618985e-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": 0.004995469911955297, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0023999007069505753, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.012390840519219636, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.01784304054453969, + "weighted_rkd_loss": 0.006195420259609818 + }, + { + "epoch": 5.419463087248322, + "flow_matching_loss": 0.013066306710243225, + "grad_norm": 0.16006220877170563, + "learning_rate": 9.165933312886944e-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": 0.004924082010984421, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.002115910965949297, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.011964075081050397, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.013066306710243225, + "weighted_rkd_loss": 0.005982037540525198 + }, + { + "epoch": 5.436241610738255, + "flow_matching_loss": 0.014801208395510912, + "grad_norm": 0.16490410268306732, + "learning_rate": 9.07466464413155e-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": 0.005378607613965869, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0020040612551383676, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.012761276494711638, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.014801208395510912, + "weighted_rkd_loss": 0.006380638247355819 + }, + { + "epoch": 5.453020134228188, + "flow_matching_loss": 0.014771063253283501, + "grad_norm": 0.25417137145996094, + "learning_rate": 8.983655168740048e-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": 0.005886884219944477, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0022363172494806348, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.014010085724294186, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.014771063253283501, + "weighted_rkd_loss": 0.007005042862147093 + }, + { + "epoch": 5.469798657718121, + "flow_matching_loss": 0.013944575842469931, + "grad_norm": 0.15955711901187897, + "learning_rate": 8.892908867761718e-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": 0.006365548493340612, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0026807011221535505, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.0154117981903255, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.013944575842469931, + "weighted_rkd_loss": 0.00770589909516275 + }, + { + "epoch": 5.4865771812080535, + "flow_matching_loss": 0.01557155018672347, + "grad_norm": 0.148858442902565, + "learning_rate": 8.802429710733745e-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": 0.004635971668176353, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0020682581351138653, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.011340201552957296, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.01557155018672347, + "weighted_rkd_loss": 0.005670100776478648 + }, + { + "epoch": 5.503355704697986, + "flow_matching_loss": 0.01386803649365902, + "grad_norm": 0.15989872813224792, + "learning_rate": 8.71222165550757e-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": 0.004535936331376433, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0017340362304821611, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.010805909056216478, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.01386803649365902, + "weighted_rkd_loss": 0.005402954528108239 + }, + { + "epoch": 5.52013422818792, + "flow_matching_loss": 0.013872923143208028, + "grad_norm": 0.20790772140026093, + "learning_rate": 8.622288648075748e-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": 0.005408620904199779, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0021506744320504366, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.012967916205525398, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.013872923143208028, + "weighted_rkd_loss": 0.006483958102762699 + }, + { + "epoch": 5.5369127516778525, + "flow_matching_loss": 0.015501438919454814, + "grad_norm": 0.15546298027038574, + "learning_rate": 8.532634622399391e-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": 0.004575241915881633, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0018669651937671006, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.011017449013888837, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.015501438919454814, + "weighted_rkd_loss": 0.005508724506944418 + }, + { + "epoch": 5.553691275167785, + "flow_matching_loss": 0.012488964200019836, + "grad_norm": 0.23867888748645782, + "learning_rate": 8.44326350023601e-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": 0.005004889238625765, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0023841552902013063, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.012393934046849608, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.012488964200019836, + "weighted_rkd_loss": 0.006196967023424804 + }, + { + "epoch": 5.570469798657718, + "flow_matching_loss": 0.013480441831052303, + "grad_norm": 0.1565876305103302, + "learning_rate": 8.354179190968002e-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": 0.004363967292010784, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.001519673701841384, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.010247608181089163, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.013480441831052303, + "weighted_rkd_loss": 0.005123804090544581 + }, + { + "epoch": 5.587248322147651, + "flow_matching_loss": 0.015796675253659485, + "grad_norm": 0.24322588741779327, + "learning_rate": 8.265385591431661e-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": 0.005845426116138697, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0023771880776621403, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01406804034486413, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.015796675253659485, + "weighted_rkd_loss": 0.007034020172432065 + }, + { + "epoch": 5.604026845637584, + "flow_matching_loss": 0.01234276294708252, + "grad_norm": 0.16083206236362457, + "learning_rate": 8.176886585746671e-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": 0.004430905706249178, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0018418193329125642, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.010703630931675433, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.01234276294708252, + "weighted_rkd_loss": 0.005351815465837717 + }, + { + "epoch": 5.620805369127517, + "flow_matching_loss": 0.01186593621969223, + "grad_norm": 0.20607060194015503, + "learning_rate": 8.088686045146247e-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": 0.004728447226807475, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.001986333087552339, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.011443227529525757, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.01186593621969223, + "weighted_rkd_loss": 0.005721613764762878 + }, + { + "epoch": 5.6375838926174495, + "flow_matching_loss": 0.013174599409103394, + "grad_norm": 0.1966661512851715, + "learning_rate": 8.00078782780775e-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": 0.004724739631637931, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.001865789678413421, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.011315269116312265, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.013174599409103394, + "weighted_rkd_loss": 0.005657634558156132 + }, + { + "epoch": 5.654362416107382, + "flow_matching_loss": 0.01585042094811797, + "grad_norm": 0.245307058095932, + "learning_rate": 7.913195778683979e-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": 0.005080026038922369, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.002161143010016531, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.012321195006370545, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.01585042094811797, + "weighted_rkd_loss": 0.006160597503185272 + }, + { + "epoch": 5.671140939597316, + "flow_matching_loss": 0.015605000220239163, + "grad_norm": 0.20907120406627655, + "learning_rate": 7.825913729334908e-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": 0.005415506870485842, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0021290503675118087, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.012960064318031072, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.015605000220239163, + "weighted_rkd_loss": 0.006480032159015536 + }, + { + "epoch": 5.6879194630872485, + "flow_matching_loss": 0.016725842375308275, + "grad_norm": 0.14447364211082458, + "learning_rate": 7.73894549776013e-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": 0.00547388365957886, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0023031249875202775, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.013250892329961062, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.016725842375308275, + "weighted_rkd_loss": 0.006625446164980531 + }, + { + "epoch": 5.704697986577181, + "flow_matching_loss": 0.01270838575437665, + "grad_norm": 0.23588861525058746, + "learning_rate": 7.65229488823184e-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": 0.005150177539326251, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0022166476352140307, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.012517002783715725, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.01270838575437665, + "weighted_rkd_loss": 0.006258501391857863 + }, + { + "epoch": 5.721476510067114, + "flow_matching_loss": 0.015314796008169651, + "grad_norm": 0.182273268699646, + "learning_rate": 7.565965691128396e-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": 0.00467446381226182, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0020193661213852463, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01136829387396574, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.015314796008169651, + "weighted_rkd_loss": 0.00568414693698287 + }, + { + "epoch": 5.7382550335570475, + "flow_matching_loss": 0.013604261912405491, + "grad_norm": 0.23861733078956604, + "learning_rate": 7.47996168276856e-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": 0.004217569250613451, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0016527582542039455, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.010087896650657058, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.013604261912405491, + "weighted_rkd_loss": 0.005043948325328529 + }, + { + "epoch": 5.75503355704698, + "flow_matching_loss": 0.014357882179319858, + "grad_norm": 0.1535225212574005, + "learning_rate": 7.394286625246258e-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": 0.0049197893822565675, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0016938321408815683, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.011533410940319299, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.014357882179319858, + "weighted_rkd_loss": 0.0057667054701596495 + }, + { + "epoch": 5.771812080536913, + "flow_matching_loss": 0.0134419372305274, + "grad_norm": 0.15331624448299408, + "learning_rate": 7.308944266266063e-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": 0.004597988841123879, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0018919462454505265, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01108792396262288, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.0134419372305274, + "weighted_rkd_loss": 0.00554396198131144 + }, + { + "epoch": 5.7885906040268456, + "flow_matching_loss": 0.014776341710239648, + "grad_norm": 0.25563931465148926, + "learning_rate": 7.223938338979233e-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": 0.004740951280109584, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0018884244142100214, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.011370326951146125, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.014776341710239648, + "weighted_rkd_loss": 0.005685163475573063 + }, + { + "epoch": 5.805369127516778, + "flow_matching_loss": 0.014994413778185844, + "grad_norm": 0.15913286805152893, + "learning_rate": 7.139272561820397e-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": 0.004294173908419907, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0015233114478178321, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.010111659159883856, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.014994413778185844, + "weighted_rkd_loss": 0.005055829579941928 + }, + { + "epoch": 5.822147651006711, + "flow_matching_loss": 0.012550541758537292, + "grad_norm": 0.171083003282547, + "learning_rate": 7.054950638344936e-06, + "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": 0.004584924993105232, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0019210978411138059, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.011090947780758142, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.012550541758537292, + "weighted_rkd_loss": 0.005545473890379071 + }, + { + "epoch": 5.8389261744966445, + "flow_matching_loss": 0.01349291019141674, + "grad_norm": 0.17222721874713898, + "learning_rate": 6.970976257066941e-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": 0.005571219907142222, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.00232730865245685, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.013469748198986053, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.01349291019141674, + "weighted_rkd_loss": 0.006734874099493027 + }, + { + "epoch": 5.855704697986577, + "flow_matching_loss": 0.013362313434481621, + "grad_norm": 0.26061469316482544, + "learning_rate": 6.887353091297896e-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": 0.005095311347395182, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0020249326713383197, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.012215555505827069, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.013362313434481621, + "weighted_rkd_loss": 0.0061077777529135345 + }, + { + "epoch": 5.87248322147651, + "flow_matching_loss": 0.014484726265072823, + "grad_norm": 0.16677652299404144, + "learning_rate": 6.804084798985965e-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": 0.0047961494885385035, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.001770413643680513, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.011362712737172842, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.014484726265072823, + "weighted_rkd_loss": 0.005681356368586421 + }, + { + "epoch": 5.889261744966443, + "flow_matching_loss": 0.015598151553422212, + "grad_norm": 0.2078307569026947, + "learning_rate": 6.721175022556009e-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": 0.005224377172999084, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0020208677859045567, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01246962221339345, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.015598151553422212, + "weighted_rkd_loss": 0.006234811106696725 + }, + { + "epoch": 5.906040268456376, + "flow_matching_loss": 0.01374701363965869, + "grad_norm": 0.2645686864852905, + "learning_rate": 6.638627388750245e-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": 0.0053967045387253165, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0019286571186967194, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.012722066324204206, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.01374701363965869, + "weighted_rkd_loss": 0.006361033162102103 + }, + { + "epoch": 5.922818791946309, + "flow_matching_loss": 0.016289659217000006, + "grad_norm": 0.2069074511528015, + "learning_rate": 6.556445508469583e-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": 0.0047795751597732306, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0018127361487131565, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.011371886404231191, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.016289659217000006, + "weighted_rkd_loss": 0.005685943202115596 + }, + { + "epoch": 5.939597315436242, + "flow_matching_loss": 0.014452755730599166, + "grad_norm": 0.1726139336824417, + "learning_rate": 6.474632976615708e-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": 0.0047163198003545405, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0022801060462370516, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.011712745716795325, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.014452755730599166, + "weighted_rkd_loss": 0.005856372858397662 + }, + { + "epoch": 5.956375838926174, + "flow_matching_loss": 0.013426850829273463, + "grad_norm": 0.1568567007780075, + "learning_rate": 6.393193371933791e-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": 0.005156427063047886, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0024364317068830132, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.012749285716563464, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.013426850829273463, + "weighted_rkd_loss": 0.006374642858281732 + }, + { + "epoch": 5.973154362416107, + "flow_matching_loss": 0.013179479539394379, + "grad_norm": 0.1591392308473587, + "learning_rate": 6.312130256855977e-06, + "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": 0.0041168633382767435, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0016634542611427606, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.009897180832922459, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.013179479539394379, + "weighted_rkd_loss": 0.004948590416461229 + }, + { + "epoch": 5.989932885906041, + "flow_matching_loss": 0.01301817810162902, + "grad_norm": 0.12511791288852692, + "learning_rate": 6.23144717734552e-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": 0.004697659728117287, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0018042239942587912, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.011199543299153448, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.01301817810162902, + "weighted_rkd_loss": 0.005599771649576724 + }, + { + "epoch": 6.006711409395973, + "flow_matching_loss": 0.01299028513021767, + "grad_norm": 0.1949714571237564, + "learning_rate": 6.151147662741695e-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": 0.004801549506373704, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.002034022466978058, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.011637121578678489, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.01299028513021767, + "weighted_rkd_loss": 0.0058185607893392445 + }, + { + "epoch": 6.023489932885906, + "flow_matching_loss": 0.014248680043965577, + "grad_norm": 0.18050147593021393, + "learning_rate": 6.071235225605415e-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": 0.004431381216272711, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0018716220394708217, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.010734384413808584, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.014248680043965577, + "weighted_rkd_loss": 0.005367192206904292 + }, + { + "epoch": 6.040268456375839, + "flow_matching_loss": 0.014908248744904994, + "grad_norm": 0.14633774757385254, + "learning_rate": 5.991713361565546e-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": 0.004304599575698375, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0017973661771975458, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.010406565340235829, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.014908248744904994, + "weighted_rkd_loss": 0.005203282670117914 + }, + { + "epoch": 6.057046979865772, + "flow_matching_loss": 0.012182641122490168, + "grad_norm": 0.231786847114563, + "learning_rate": 5.912585549166051e-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": 0.0053547847550362345, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0023295242921449245, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.013039093744009734, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.012182641122490168, + "weighted_rkd_loss": 0.006519546872004867 + }, + { + "epoch": 6.073825503355705, + "flow_matching_loss": 0.01349649541079998, + "grad_norm": 0.15987418591976166, + "learning_rate": 5.833855249713777e-06, + "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": 0.0037824744707904757, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0015456155641004443, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.009110564505681396, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.01349649541079998, + "weighted_rkd_loss": 0.004555282252840698 + }, + { + "epoch": 6.090604026845638, + "flow_matching_loss": 0.014953711070120335, + "grad_norm": 0.17302507162094116, + "learning_rate": 5.7555259071270645e-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": 0.004880516487173736, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.001975899317767471, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.011736932350322605, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.014953711070120335, + "weighted_rkd_loss": 0.005868466175161302 + }, + { + "epoch": 6.10738255033557, + "flow_matching_loss": 0.012672061333432794, + "grad_norm": 0.1585019826889038, + "learning_rate": 5.677600947785117e-06, + "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": 0.0042997630778700115, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0017347801825962962, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.010334306443110109, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.012672061333432794, + "weighted_rkd_loss": 0.0051671532215550545 + }, + { + "epoch": 6.124161073825503, + "flow_matching_loss": 0.013576794881373644, + "grad_norm": 0.1438910812139511, + "learning_rate": 5.600083780378099e-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": 0.0039490525843575595, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.001519838348031044, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.009417943563312292, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.013576794881373644, + "weighted_rkd_loss": 0.004708971781656146 + }, + { + "epoch": 6.140939597315437, + "flow_matching_loss": 0.013494531717151403, + "grad_norm": 0.19352352619171143, + "learning_rate": 5.522977795758038e-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": 0.004718074877746403, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0016840157913975418, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.011120165511965752, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.013494531717151403, + "weighted_rkd_loss": 0.005560082755982876 + }, + { + "epoch": 6.157718120805369, + "flow_matching_loss": 0.014294523280113935, + "grad_norm": 0.1627226024866104, + "learning_rate": 5.4462863667904805e-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": 0.00501574594527483, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0020766448462381957, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.012108136806637049, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.014294523280113935, + "weighted_rkd_loss": 0.0060540684033185245 + }, + { + "epoch": 6.174496644295302, + "flow_matching_loss": 0.015006358921527862, + "grad_norm": 0.18388888239860535, + "learning_rate": 5.370012848206983e-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": 0.005103649152442813, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.002023331017699093, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.012230629660189151, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.015006358921527862, + "weighted_rkd_loss": 0.006115314830094576 + }, + { + "epoch": 6.191275167785235, + "flow_matching_loss": 0.01274308622814715, + "grad_norm": 0.19488470256328583, + "learning_rate": 5.29416057645834e-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": 0.004091479768976569, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0017668359796516597, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.00994979557581246, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.01274308622814715, + "weighted_rkd_loss": 0.00497489778790623 + }, + { + "epoch": 6.208053691275167, + "flow_matching_loss": 0.013387705152854323, + "grad_norm": 0.1380261331796646, + "learning_rate": 5.218732869568635e-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": 0.004827187396585941, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0020433113211765887, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.011697686091065407, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.013387705152854323, + "weighted_rkd_loss": 0.005848843045532704 + }, + { + "epoch": 6.224832214765101, + "flow_matching_loss": 0.011581274913623929, + "grad_norm": 0.15014977753162384, + "learning_rate": 5.14373302699013e-06, + "loss": 0.0167, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.004214526782743633, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0017126499151345342, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.010141703393310308, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.011581274913623929, + "weighted_rkd_loss": 0.005070851696655154 + }, + { + "epoch": 6.241610738255034, + "flow_matching_loss": 0.013908269163221122, + "grad_norm": 0.1939009577035904, + "learning_rate": 5.069164329458894e-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": 0.0045234568417072294, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0020099034416489305, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.011056817043572665, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.013908269163221122, + "weighted_rkd_loss": 0.005528408521786332 + }, + { + "epoch": 6.258389261744966, + "flow_matching_loss": 0.014544932637363672, + "grad_norm": 0.20686961710453033, + "learning_rate": 4.995030038851342e-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": 0.0044759223237633705, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0015815795748494565, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.010533424327149987, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.014544932637363672, + "weighted_rkd_loss": 0.005266712163574994 + }, + { + "epoch": 6.275167785234899, + "flow_matching_loss": 0.012062178738415242, + "grad_norm": 0.141107439994812, + "learning_rate": 4.9213333980415e-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": 0.0045945595251396295, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0018156210309825838, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.011004740232601761, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.012062178738415242, + "weighted_rkd_loss": 0.005502370116300881 + }, + { + "epoch": 6.291946308724833, + "flow_matching_loss": 0.013018049858510494, + "grad_norm": 0.147357776761055, + "learning_rate": 4.848077630759175e-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": 0.004304882767610252, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0018924566567875445, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.010502222133800388, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.013018049858510494, + "weighted_rkd_loss": 0.005251111066900194 + }, + { + "epoch": 6.308724832214765, + "flow_matching_loss": 0.014812441915273667, + "grad_norm": 0.21773049235343933, + "learning_rate": 4.775265941448951e-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": 0.004822571738623082, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0017652085982263089, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.011410352075472474, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.014812441915273667, + "weighted_rkd_loss": 0.005705176037736237 + }, + { + "epoch": 6.325503355704698, + "flow_matching_loss": 0.013156043365597725, + "grad_norm": 0.15065298974514008, + "learning_rate": 4.702901515129989e-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": 0.004030594695359468, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0016236922470852732, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.009684881567955017, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.013156043365597725, + "weighted_rkd_loss": 0.004842440783977509 + }, + { + "epoch": 6.342281879194631, + "flow_matching_loss": 0.01194716515019536, + "grad_norm": 0.18975025415420532, + "learning_rate": 4.630987517256734e-06, + "loss": 0.0164, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.0037176207872107623, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.001498365867882967, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.008933607395738364, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.01194716515019536, + "weighted_rkd_loss": 0.004466803697869182 + }, + { + "epoch": 6.359060402684563, + "flow_matching_loss": 0.013762779394164681, + "grad_norm": 0.17298924922943115, + "learning_rate": 4.559527093580412e-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": 0.004396662302315235, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0016519619966857135, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.010445286706089973, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.013762779394164681, + "weighted_rkd_loss": 0.005222643353044986 + }, + { + "epoch": 6.375838926174497, + "flow_matching_loss": 0.013622605893760919, + "grad_norm": 0.13369767367839813, + "learning_rate": 4.488523370011456e-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": 0.003639714792370796, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0014581867260858416, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.008737616240978241, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.013622605893760919, + "weighted_rkd_loss": 0.0043688081204891205 + }, + { + "epoch": 6.39261744966443, + "flow_matching_loss": 0.012390201818197966, + "grad_norm": 0.13894836604595184, + "learning_rate": 4.417979452482766e-06, + "loss": 0.017, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.003824461461044848, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.001623190485406667, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.00927211344242096, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.012390201818197966, + "weighted_rkd_loss": 0.00463605672121048 + }, + { + "epoch": 6.409395973154362, + "flow_matching_loss": 0.011436420166864992, + "grad_norm": 0.13240008056163788, + "learning_rate": 4.347898426813811e-06, + "loss": 0.0172, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.004742332524619997, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.002084954851306975, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01156962001696229, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.011436420166864992, + "weighted_rkd_loss": 0.005784810008481145 + }, + { + "epoch": 6.426174496644295, + "flow_matching_loss": 0.012280825292691588, + "grad_norm": 0.14240750670433044, + "learning_rate": 4.278283358575694e-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": 0.00448585560079664, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0018520746380090714, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.010823785979300737, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.012280825292691588, + "weighted_rkd_loss": 0.005411892989650368 + }, + { + "epoch": 6.442953020134228, + "flow_matching_loss": 0.012999581918120385, + "grad_norm": 0.14199872314929962, + "learning_rate": 4.2091372929570145e-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": 0.004581688530743122, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0020241894177161156, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.011187566444277763, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.012999581918120385, + "weighted_rkd_loss": 0.005593783222138882 + }, + { + "epoch": 6.459731543624161, + "flow_matching_loss": 0.015608004108071328, + "grad_norm": 0.17531105875968933, + "learning_rate": 4.1404632546306934e-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": 0.004717425419948995, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0018084937881212681, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01124334461055696, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.015608004108071328, + "weighted_rkd_loss": 0.00562167230527848 + }, + { + "epoch": 6.476510067114094, + "flow_matching_loss": 0.01236470346339047, + "grad_norm": 0.19072523713111877, + "learning_rate": 4.0722642476216286e-06, + "loss": 0.0171, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.003976859035901725, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.001551550440490246, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.009505268465727568, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.01236470346339047, + "weighted_rkd_loss": 0.004752634232863784 + }, + { + "epoch": 6.493288590604027, + "flow_matching_loss": 0.013742083543911576, + "grad_norm": 0.13357122242450714, + "learning_rate": 4.00454325517533e-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": 0.004746155813336372, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0017761539667844773, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.011268465546891093, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.013742083543911576, + "weighted_rkd_loss": 0.005634232773445547 + }, + { + "epoch": 6.510067114093959, + "flow_matching_loss": 0.01476223086938262, + "grad_norm": 0.2286960482597351, + "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": 0.004158663819544018, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0018114029197022318, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.010128730582073331, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.01476223086938262, + "weighted_rkd_loss": 0.005064365291036666 + }, + { + "epoch": 6.526845637583893, + "flow_matching_loss": 0.013768768729642033, + "grad_norm": 0.36107489466667175, + "learning_rate": 3.870547142273918e-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": 0.005147815262898803, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0018881296273320913, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.012183760106563569, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.013768768729642033, + "weighted_rkd_loss": 0.006091880053281784 + }, + { + "epoch": 6.543624161073826, + "flow_matching_loss": 0.013367434125393629, + "grad_norm": 0.2914619743824005, + "learning_rate": 3.8042778832428736e-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": 0.004395952657796442, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0020264136604964734, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.010818319115787744, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.013367434125393629, + "weighted_rkd_loss": 0.005409159557893872 + }, + { + "epoch": 6.560402684563758, + "flow_matching_loss": 0.017430961970239876, + "grad_norm": 0.159344881772995, + "learning_rate": 3.7384983613663195e-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": 0.0042855771724134685, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0016469669179059565, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.010218121297657489, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.017430961970239876, + "weighted_rkd_loss": 0.0051090606488287445 + }, + { + "epoch": 6.577181208053691, + "flow_matching_loss": 0.014220637455582618, + "grad_norm": 0.1419205665588379, + "learning_rate": 3.6732114540536503e-06, + "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": 0.003935556556098163, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.001527376810554415, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.009398489771410824, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.014220637455582618, + "weighted_rkd_loss": 0.004699244885705412 + }, + { + "epoch": 6.593959731543624, + "flow_matching_loss": 0.013942512311041355, + "grad_norm": 0.18272189795970917, + "learning_rate": 3.608420017165683e-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": 0.0038367455126717687, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.001534297107718885, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.0092077880166471, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.013942512311041355, + "weighted_rkd_loss": 0.00460389400832355 + }, + { + "epoch": 6.610738255033557, + "flow_matching_loss": 0.015249136649072171, + "grad_norm": 0.16941304504871368, + "learning_rate": 3.5441268848897664e-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": 0.00395207954570651, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0015867861919105053, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.009490945329889654, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.015249136649072171, + "weighted_rkd_loss": 0.004745472664944827 + }, + { + "epoch": 6.62751677852349, + "flow_matching_loss": 0.012341193947941064, + "grad_norm": 0.1718492954969406, + "learning_rate": 3.480334869615792e-06, + "loss": 0.0173, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.004190959827974438, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0015963940415531397, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.009978313744068146, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.012341193947941064, + "weighted_rkd_loss": 0.004989156872034073 + }, + { + "epoch": 6.644295302013423, + "flow_matching_loss": 0.013152943830937147, + "grad_norm": 0.16254186630249023, + "learning_rate": 3.417046761813155e-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": 0.004420910519547761, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0017289884388446807, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.010570809524506331, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.013152943830937147, + "weighted_rkd_loss": 0.005285404762253166 + }, + { + "epoch": 6.6610738255033555, + "flow_matching_loss": 0.01244001667946577, + "grad_norm": 0.13000580668449402, + "learning_rate": 3.3542653299087263e-06, + "loss": 0.0173, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.004057411407120526, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0015844591427594424, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.009699281957000493, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.01244001667946577, + "weighted_rkd_loss": 0.004849640978500247 + }, + { + "epoch": 6.677852348993289, + "flow_matching_loss": 0.010493615129962563, + "grad_norm": 0.15268608927726746, + "learning_rate": 3.2919933201657117e-06, + "loss": 0.0159, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.004557304014451802, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.001655701722484082, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01077030966989696, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.010493615129962563, + "weighted_rkd_loss": 0.00538515483494848 + }, + { + "epoch": 6.694630872483222, + "flow_matching_loss": 0.013269064854830504, + "grad_norm": 0.18566147983074188, + "learning_rate": 3.2302334565635584e-06, + "loss": 0.0179, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.003922327677719295, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0013941960758529603, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.009238851442933082, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.013269064854830504, + "weighted_rkd_loss": 0.004619425721466541 + }, + { + "epoch": 6.7114093959731544, + "flow_matching_loss": 0.012347688805311919, + "grad_norm": 0.1489165872335434, + "learning_rate": 3.1689884406787674e-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": 0.004499934636987746, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0015349635505117476, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.010534832859411836, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.012347688805311919, + "weighted_rkd_loss": 0.005267416429705918 + }, + { + "epoch": 6.728187919463087, + "flow_matching_loss": 0.012257662788033485, + "grad_norm": 0.1501564085483551, + "learning_rate": 3.1082609515667424e-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": 0.005181431001983583, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.001985070831142366, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.012347932904958725, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.012257662788033485, + "weighted_rkd_loss": 0.0061739664524793625 + }, + { + "epoch": 6.74496644295302, + "flow_matching_loss": 0.014351728092879056, + "grad_norm": 0.17460209131240845, + "learning_rate": 3.048053645644594e-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": 0.004749392857775092, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0019455971429124474, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.011444382835179567, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.014351728092879056, + "weighted_rkd_loss": 0.005722191417589784 + }, + { + "epoch": 6.7617449664429525, + "flow_matching_loss": 0.012328914739191532, + "grad_norm": 0.18522049486637115, + "learning_rate": 2.988369156574916e-06, + "loss": 0.0169, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.0039043401600793006, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0013507586263585836, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.00915943905711174, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.012328914739191532, + "weighted_rkd_loss": 0.00457971952855587 + }, + { + "epoch": 6.778523489932886, + "flow_matching_loss": 0.012781216204166413, + "grad_norm": 0.15217529237270355, + "learning_rate": 2.9292100951506225e-06, + "loss": 0.0168, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.003358302265405655, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0013781144050881266, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.008094718912616372, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.012781216204166413, + "weighted_rkd_loss": 0.004047359456308186 + }, + { + "epoch": 6.795302013422819, + "flow_matching_loss": 0.012864326685667038, + "grad_norm": 0.19025535881519318, + "learning_rate": 2.8705790491807033e-06, + "loss": 0.0173, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.003772235964424908, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.001426016999175772, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.008970488933846355, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.012864326685667038, + "weighted_rkd_loss": 0.004485244466923177 + }, + { + "epoch": 6.8120805369127515, + "flow_matching_loss": 0.012307855812832714, + "grad_norm": 0.19388285279273987, + "learning_rate": 2.8124785833770433e-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": 0.004534709919244051, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.001737466419581324, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.010806886246427894, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.012307855812832714, + "weighted_rkd_loss": 0.005403443123213947 + }, + { + "epoch": 6.828859060402684, + "flow_matching_loss": 0.015203884430229664, + "grad_norm": 0.15002429485321045, + "learning_rate": 2.754911239242241e-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": 0.003685287549160421, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0016135861864313484, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.008984161168336868, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.015203884430229664, + "weighted_rkd_loss": 0.004492080584168434 + }, + { + "epoch": 6.845637583892618, + "flow_matching_loss": 0.012085525365546346, + "grad_norm": 0.11901740729808807, + "learning_rate": 2.6978795349584204e-06, + "loss": 0.0159, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.0031661784276366236, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0012743374507408589, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.007606694335117936, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.012085525365546346, + "weighted_rkd_loss": 0.003803347167558968 + }, + { + "epoch": 6.8624161073825505, + "flow_matching_loss": 0.01438704738393426, + "grad_norm": 0.16812318563461304, + "learning_rate": 2.641385965277088e-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": 0.003867070819251239, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0017184988828375936, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.009452640358358622, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.01438704738393426, + "weighted_rkd_loss": 0.004726320179179311 + }, + { + "epoch": 6.879194630872483, + "flow_matching_loss": 0.012696548271924258, + "grad_norm": 0.14939044415950775, + "learning_rate": 2.5854330014099923e-06, + "loss": 0.0171, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.0036327670561149715, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.001501097227446735, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.008766631409525872, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.012696548271924258, + "weighted_rkd_loss": 0.004383315704762936 + }, + { + "epoch": 6.895973154362416, + "flow_matching_loss": 0.013077494082972407, + "grad_norm": 0.17584958672523499, + "learning_rate": 2.5300230909210298e-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": 0.004372379346750677, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.001655038306489587, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01039979699999094, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.013077494082972407, + "weighted_rkd_loss": 0.00519989849999547 + }, + { + "epoch": 6.912751677852349, + "flow_matching_loss": 0.01232621893286705, + "grad_norm": 0.14323139190673828, + "learning_rate": 2.4751586576191983e-06, + "loss": 0.0168, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.003784377779811621, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0013999658403918147, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.008968721283599735, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.01232621893286705, + "weighted_rkd_loss": 0.004484360641799867 + }, + { + "epoch": 6.929530201342282, + "flow_matching_loss": 0.011209582397714257, + "grad_norm": 0.15581539273262024, + "learning_rate": 2.420842101452538e-06, + "loss": 0.0167, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.004541526525281369, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.001841145200887695, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.010924198292195797, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.011209582397714257, + "weighted_rkd_loss": 0.005462099146097898 + }, + { + "epoch": 6.946308724832215, + "flow_matching_loss": 0.011785591254010797, + "grad_norm": 0.16155187785625458, + "learning_rate": 2.3670757984031816e-06, + "loss": 0.016, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.0035794392693787812, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0013535585603676736, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.008512437157332898, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.011785591254010797, + "weighted_rkd_loss": 0.004256218578666449 + }, + { + "epoch": 6.9630872483221475, + "flow_matching_loss": 0.010752106131985784, + "grad_norm": 0.2019570916891098, + "learning_rate": 2.313862100383404e-06, + "loss": 0.0158, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.004264209792017937, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0015931188769172876, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.010121538490056991, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.010752106131985784, + "weighted_rkd_loss": 0.0050607692450284954 + }, + { + "epoch": 6.97986577181208, + "flow_matching_loss": 0.010426551662385463, + "grad_norm": 0.1295207440853119, + "learning_rate": 2.2612033351327406e-06, + "loss": 0.0148, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.003634925140067935, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0014228249609004706, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.008692675270140171, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.010426551662385463, + "weighted_rkd_loss": 0.0043463376350700855 + }, + { + "epoch": 6.996644295302014, + "flow_matching_loss": 0.012410965654999018, + "grad_norm": 0.15585634112358093, + "learning_rate": 2.2091018061161788e-06, + "loss": 0.0171, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.0038361755898222325, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0016191505710594356, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.009291501902043819, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.012410965654999018, + "weighted_rkd_loss": 0.004645750951021909 + }, + { + "epoch": 7.0134228187919465, + "flow_matching_loss": 0.013164480961859227, + "grad_norm": 0.16433636844158173, + "learning_rate": 2.15755979242338e-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": 0.003910991130396724, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0018357451888732612, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.009657727554440498, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.013164480961859227, + "weighted_rkd_loss": 0.004828863777220249 + }, + { + "epoch": 7.030201342281879, + "flow_matching_loss": 0.013598700612783432, + "grad_norm": 0.16424773633480072, + "learning_rate": 2.106579548669e-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": 0.004430685797706247, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0016169998794794082, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.010478371428325773, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.013598700612783432, + "weighted_rkd_loss": 0.005239185714162886 + }, + { + "epoch": 7.046979865771812, + "flow_matching_loss": 0.011333533329889178, + "grad_norm": 0.15079383552074432, + "learning_rate": 2.0561633048940477e-06, + "loss": 0.0161, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.003885891393292695, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0017129390267655253, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.009484721766784786, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.011333533329889178, + "weighted_rkd_loss": 0.004742360883392393 + }, + { + "epoch": 7.063758389261745, + "flow_matching_loss": 0.013496446702629328, + "grad_norm": 0.14999589323997498, + "learning_rate": 2.006313266468355e-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": 0.004368492448702454, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0016905601136386395, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.010427544871345163, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.013496446702629328, + "weighted_rkd_loss": 0.0052137724356725815 + }, + { + "epoch": 7.080536912751678, + "flow_matching_loss": 0.012888733576983214, + "grad_norm": 0.18489472568035126, + "learning_rate": 1.957031613994097e-06, + "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": 0.0038338485173881054, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.001534275064477697, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.00920197213999927, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.012888733576983214, + "weighted_rkd_loss": 0.004600986069999635 + }, + { + "epoch": 7.097315436241611, + "flow_matching_loss": 0.012950423266738653, + "grad_norm": 0.13276706635951996, + "learning_rate": 1.9083205032103995e-06, + "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": 0.0038131119683384894, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0015088815591298044, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.009135105554014445, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.012950423266738653, + "weighted_rkd_loss": 0.004567552777007222 + }, + { + "epoch": 7.114093959731544, + "flow_matching_loss": 0.013198824971914292, + "grad_norm": 0.1334235966205597, + "learning_rate": 1.8601820648990536e-06, + "loss": 0.0174, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.003466808726079762, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0014216269366443158, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.008355244342237712, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.013198824971914292, + "weighted_rkd_loss": 0.004177622171118856 + }, + { + "epoch": 7.130872483221476, + "flow_matching_loss": 0.012214665487408637, + "grad_norm": 0.17523948848247528, + "learning_rate": 1.8126184047912936e-06, + "loss": 0.0163, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.003384548705071211, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0013439524045679718, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.008113049762323498, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.012214665487408637, + "weighted_rkd_loss": 0.004056524881161749 + }, + { + "epoch": 7.14765100671141, + "flow_matching_loss": 0.011672427924349905, + "grad_norm": 0.12189946323633194, + "learning_rate": 1.7656316034757024e-06, + "loss": 0.0162, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.0036959612276405095, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0017431621032301336, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.00913508445955813, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.011672427924349905, + "weighted_rkd_loss": 0.004567542229779065 + }, + { + "epoch": 7.1644295302013425, + "flow_matching_loss": 0.012925780657678843, + "grad_norm": 0.17717908322811127, + "learning_rate": 1.7192237163071816e-06, + "loss": 0.0171, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.003533428453374654, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0013367140141781419, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.008403570856899023, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.012925780657678843, + "weighted_rkd_loss": 0.0042017854284495115 + }, + { + "epoch": 7.181208053691275, + "flow_matching_loss": 0.013103020470589399, + "grad_norm": 0.11885635554790497, + "learning_rate": 1.6733967733170573e-06, + "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": 0.003977324068546295, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0015103903540875762, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.00946503859013319, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.013103020470589399, + "weighted_rkd_loss": 0.004732519295066595 + }, + { + "epoch": 7.197986577181208, + "flow_matching_loss": 0.010397190414369106, + "grad_norm": 0.16860656440258026, + "learning_rate": 1.628152779124279e-06, + "loss": 0.0151, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.0038975286413915457, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0016264894686173648, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.009421546664088964, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.010397190414369106, + "weighted_rkd_loss": 0.004710773332044482 + }, + { + "epoch": 7.214765100671141, + "flow_matching_loss": 0.012783728912472725, + "grad_norm": 0.12303238362073898, + "learning_rate": 1.583493712847715e-06, + "loss": 0.0174, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.003879416105337441, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0015254201600328088, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.009284252347424627, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.012783728912472725, + "weighted_rkd_loss": 0.0046421261737123135 + }, + { + "epoch": 7.231543624161074, + "flow_matching_loss": 0.013865695055574178, + "grad_norm": 0.1526777744293213, + "learning_rate": 1.5394215280196084e-06, + "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": 0.003514062752947211, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0013878538622520864, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.00841597942635417, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.013865695055574178, + "weighted_rkd_loss": 0.004207989713177085 + }, + { + "epoch": 7.248322147651007, + "flow_matching_loss": 0.013047731202095746, + "grad_norm": 0.16690459847450256, + "learning_rate": 1.4959381525000904e-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": 0.004131821263581514, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0015723746037110686, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.009836017154157162, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.013047731202095746, + "weighted_rkd_loss": 0.004918008577078581 + }, + { + "epoch": 7.26510067114094, + "flow_matching_loss": 0.013816094864159823, + "grad_norm": 0.1267736554145813, + "learning_rate": 1.453045488392879e-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": 0.003801518469117582, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0012555589317344129, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.008858596067875623, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.013816094864159823, + "weighted_rkd_loss": 0.004429298033937812 + }, + { + "epoch": 7.281879194630872, + "flow_matching_loss": 0.012940576393157243, + "grad_norm": 0.13604986667633057, + "learning_rate": 1.4107454119620484e-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": 0.004049134370870888, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0014924665214493872, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.009590735286474228, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.012940576393157243, + "weighted_rkd_loss": 0.004795367643237114 + }, + { + "epoch": 7.298657718120805, + "flow_matching_loss": 0.013075377326458693, + "grad_norm": 0.18737582862377167, + "learning_rate": 1.3690397735499778e-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": 0.004112353594973683, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0015578099992126226, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.009782517235726118, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.013075377326458693, + "weighted_rkd_loss": 0.004891258617863059 + }, + { + "epoch": 7.315436241610739, + "flow_matching_loss": 0.013750377995893358, + "grad_norm": 0.19063593447208405, + "learning_rate": 1.327930397496403e-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": 0.003749578760471195, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0014163357904180885, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.008915493311360479, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.013750377995893358, + "weighted_rkd_loss": 0.0044577466556802395 + }, + { + "epoch": 7.332214765100671, + "flow_matching_loss": 0.010600294126197696, + "grad_norm": 0.12065929174423218, + "learning_rate": 1.287419082058604e-06, + "loss": 0.0151, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.00381252421066165, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.001365220220759511, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.00899026868864894, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.010600294126197696, + "weighted_rkd_loss": 0.00449513434432447 + }, + { + "epoch": 7.348993288590604, + "flow_matching_loss": 0.011868054792284966, + "grad_norm": 0.13562294840812683, + "learning_rate": 1.2475075993327589e-06, + "loss": 0.0162, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.003603435377590358, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0015282828360795974, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.008735153544694185, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.011868054792284966, + "weighted_rkd_loss": 0.004367576772347092 + }, + { + "epoch": 7.365771812080537, + "flow_matching_loss": 0.011605020193383098, + "grad_norm": 0.1373264491558075, + "learning_rate": 1.208197695176413e-06, + "loss": 0.0162, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.0038801249815151096, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0015233642479870468, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.009283614344894886, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.011605020193383098, + "weighted_rkd_loss": 0.004641807172447443 + }, + { + "epoch": 7.382550335570469, + "flow_matching_loss": 0.01095195272937417, + "grad_norm": 0.1840600073337555, + "learning_rate": 1.1694910891321259e-06, + "loss": 0.0158, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.004077260848134756, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0014926740492228418, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.009647195786237716, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.01095195272937417, + "weighted_rkd_loss": 0.004823597893118858 + }, + { + "epoch": 7.399328859060403, + "flow_matching_loss": 0.012686336040496826, + "grad_norm": 0.24466432631015778, + "learning_rate": 1.1313894743522363e-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": 0.004122675489634275, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0017858694423921405, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.010031220503151417, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.012686336040496826, + "weighted_rkd_loss": 0.005015610251575708 + }, + { + "epoch": 7.416107382550336, + "flow_matching_loss": 0.012575617153197527, + "grad_norm": 0.17857523262500763, + "learning_rate": 1.0938945175248016e-06, + "loss": 0.0169, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.003530106134712696, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0015555102261714637, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.008615722553804517, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.012575617153197527, + "weighted_rkd_loss": 0.004307861276902259 + }, + { + "epoch": 7.432885906040268, + "flow_matching_loss": 0.014236471243202686, + "grad_norm": 0.11501364409923553, + "learning_rate": 1.0570078588007086e-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": 0.0043713633203879, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0016129569092299789, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.010355683555826545, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.014236471243202686, + "weighted_rkd_loss": 0.005177841777913273 + }, + { + "epoch": 7.449664429530201, + "flow_matching_loss": 0.01142531242221594, + "grad_norm": 0.19348150491714478, + "learning_rate": 1.0207311117219009e-06, + "loss": 0.0167, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.00433938808273524, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0017794008366763593, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.010458177095279098, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.01142531242221594, + "weighted_rkd_loss": 0.005229088547639549 + }, + { + "epoch": 7.466442953020135, + "flow_matching_loss": 0.011046114470809698, + "grad_norm": 0.17002864181995392, + "learning_rate": 9.850658631508198e-07, + "loss": 0.0155, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.003723083692602813, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0014892871375195682, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.008935454487800597, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.011046114470809698, + "weighted_rkd_loss": 0.004467727243900299 + }, + { + "epoch": 7.483221476510067, + "flow_matching_loss": 0.012161614559590816, + "grad_norm": 0.15816764533519745, + "learning_rate": 9.500136732009751e-07, + "loss": 0.0172, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.004277206980623305, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0014976980164647102, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.010052111838012934, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.012161614559590816, + "weighted_rkd_loss": 0.005026055919006467 + }, + { + "epoch": 7.5, + "flow_matching_loss": 0.011525012785568833, + "grad_norm": 0.1495368480682373, + "learning_rate": 9.155760751687054e-07, + "loss": 0.0158, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.0035916498396545648, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0013772728387266397, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.00856057251803577, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.011525012785568833, + "weighted_rkd_loss": 0.004280286259017885 + }, + { + "epoch": 7.516778523489933, + "flow_matching_loss": 0.011997128324583172, + "grad_norm": 0.1800948828458786, + "learning_rate": 8.817545754661155e-07, + "loss": 0.017, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.004208199260756373, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0015775962849147618, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.009993994794785976, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.011997128324583172, + "weighted_rkd_loss": 0.004996997397392988 + }, + { + "epoch": 7.533557046979865, + "flow_matching_loss": 0.011707661580294371, + "grad_norm": 0.18044085800647736, + "learning_rate": 8.485506535551668e-07, + "loss": 0.0163, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.003826503362506628, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0015013229451142252, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.009154329635202884, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.011707661580294371, + "weighted_rkd_loss": 0.004577164817601442 + }, + { + "epoch": 7.550335570469799, + "flow_matching_loss": 0.011094614583998918, + "grad_norm": 0.1431879699230194, + "learning_rate": 8.159657618829685e-07, + "loss": 0.0154, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.0035614509135484694, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0014412890886887908, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.008564190845936537, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.011094614583998918, + "weighted_rkd_loss": 0.004282095422968269 + }, + { + "epoch": 7.567114093959732, + "flow_matching_loss": 0.01384103624150157, + "grad_norm": 0.1595686376094818, + "learning_rate": 7.840013258182471e-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": 0.003950062114745379, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0015118620882276445, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.009411986311897635, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.01384103624150157, + "weighted_rkd_loss": 0.0047059931559488176 + }, + { + "epoch": 7.583892617449664, + "flow_matching_loss": 0.01329559963196516, + "grad_norm": 0.17983092367649078, + "learning_rate": 7.526587435889809e-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": 0.004146448709070683, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.001896186772501096, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.010189084196463227, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.01329559963196516, + "weighted_rkd_loss": 0.0050945420982316135 + }, + { + "epoch": 7.600671140939597, + "flow_matching_loss": 0.012888224283233284, + "grad_norm": 0.1517045646905899, + "learning_rate": 7.219393862212593e-07, + "loss": 0.0174, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.0037810103734955192, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0014952955476474016, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.0090573163703084, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.012888224283233284, + "weighted_rkd_loss": 0.0045286581851542 + }, + { + "epoch": 7.617449664429531, + "flow_matching_loss": 0.01240743980742991, + "grad_norm": 0.19681622087955475, + "learning_rate": 6.918445974792819e-07, + "loss": 0.0173, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.004042636440135539, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0017130049236584455, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.009798277774825693, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.01240743980742991, + "weighted_rkd_loss": 0.004899138887412846 + }, + { + "epoch": 7.634228187919463, + "flow_matching_loss": 0.012322596600279212, + "grad_norm": 0.1485264003276825, + "learning_rate": 6.6237569380661e-07, + "loss": 0.0167, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.003632089029997587, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0015278115111868828, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.00879198950715363, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.012322596600279212, + "weighted_rkd_loss": 0.004395994753576815 + }, + { + "epoch": 7.651006711409396, + "flow_matching_loss": 0.012507512792944908, + "grad_norm": 0.21264411509037018, + "learning_rate": 6.335339642685495e-07, + "loss": 0.0174, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.004084891243837774, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.001712805643910542, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.009882588125765323, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.012507512792944908, + "weighted_rkd_loss": 0.0049412940628826615 + }, + { + "epoch": 7.667785234899329, + "flow_matching_loss": 0.011168187530711294, + "grad_norm": 0.16030879318714142, + "learning_rate": 6.05320670495787e-07, + "loss": 0.0154, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.0035492916591465472, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0014431263203732668, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.008541709743440151, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.011168187530711294, + "weighted_rkd_loss": 0.004270854871720076 + }, + { + "epoch": 7.684563758389261, + "flow_matching_loss": 0.012330411188304424, + "grad_norm": 0.13624970614910126, + "learning_rate": 5.777370466291954e-07, + "loss": 0.0168, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.003753046621568501, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0015292260563001038, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.00903531932272017, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.012330411188304424, + "weighted_rkd_loss": 0.004517659661360085 + }, + { + "epoch": 7.701342281879195, + "flow_matching_loss": 0.011879453249275685, + "grad_norm": 0.14977416396141052, + "learning_rate": 5.507842992658336e-07, + "loss": 0.0155, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.0029674590798094867, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0013971363950986415, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.007332054572179914, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.011879453249275685, + "weighted_rkd_loss": 0.003666027286089957 + }, + { + "epoch": 7.718120805369128, + "flow_matching_loss": 0.013211170118302107, + "grad_norm": 0.153207927942276, + "learning_rate": 5.24463607406192e-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": 0.003935256344266236, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0013785569346509874, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.00924906968139112, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.013211170118302107, + "weighted_rkd_loss": 0.00462453484069556 + }, + { + "epoch": 7.73489932885906, + "flow_matching_loss": 0.011284657474607229, + "grad_norm": 0.1558067351579666, + "learning_rate": 4.987761224025944e-07, + "loss": 0.0163, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.004252262297086418, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.001456828135997057, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.00996135277673602, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.011284657474607229, + "weighted_rkd_loss": 0.00498067638836801 + }, + { + "epoch": 7.751677852348993, + "flow_matching_loss": 0.015760020352900028, + "grad_norm": 0.15146303176879883, + "learning_rate": 4.737229679088556e-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": 0.0038813820108771326, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0013940333214122803, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.009156797314062715, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.015760020352900028, + "weighted_rkd_loss": 0.004578398657031357 + }, + { + "epoch": 7.768456375838926, + "flow_matching_loss": 0.01323346383869648, + "grad_norm": 0.18961799144744873, + "learning_rate": 4.493052398311115e-07, + "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": 0.0036372275091707705, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.001513058808632195, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.008787513757124544, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.01323346383869648, + "weighted_rkd_loss": 0.004393756878562272 + }, + { + "epoch": 7.785234899328859, + "flow_matching_loss": 0.012718648929148913, + "grad_norm": 0.13715103268623352, + "learning_rate": 4.2552400627989043e-07, + "loss": 0.0172, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.003695989469997585, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0016332714585587382, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.009025250328704715, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.012718648929148913, + "weighted_rkd_loss": 0.004512625164352357 + }, + { + "epoch": 7.802013422818792, + "flow_matching_loss": 0.013017370831221342, + "grad_norm": 0.15020433068275452, + "learning_rate": 4.023803075233923e-07, + "loss": 0.0169, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.0032229928066954015, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0014095393009483815, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.007855524914339185, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.013017370831221342, + "weighted_rkd_loss": 0.0039277624571695926 + }, + { + "epoch": 7.818791946308725, + "flow_matching_loss": 0.013933033216744661, + "grad_norm": 0.11856479197740555, + "learning_rate": 3.7987515594196997e-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": 0.004223480657674372, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.001563948963303119, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.010010910406708717, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.013933033216744661, + "weighted_rkd_loss": 0.005005455203354358 + }, + { + "epoch": 7.8355704697986575, + "flow_matching_loss": 0.01209373315796256, + "grad_norm": 0.16292829811573029, + "learning_rate": 3.580095359838648e-07, + "loss": 0.0162, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.0034446867881342768, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0013271984236780554, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.008216572040691972, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.01209373315796256, + "weighted_rkd_loss": 0.004108286020345986 + }, + { + "epoch": 7.85234899328859, + "flow_matching_loss": 0.011257783556357027, + "grad_norm": 0.12844575941562653, + "learning_rate": 3.3678440412212576e-07, + "loss": 0.0151, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.0032084846636280416, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0011872378352563828, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.007604207145050168, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.011257783556357027, + "weighted_rkd_loss": 0.003802103572525084 + }, + { + "epoch": 7.869127516778524, + "flow_matching_loss": 0.013048629881814123, + "grad_norm": 0.1467456817626953, + "learning_rate": 3.162006888127833e-07, + "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": 0.004155136458575725, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0016613823710940779, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.009971655253320933, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.013048629881814123, + "weighted_rkd_loss": 0.0049858276266604665 + }, + { + "epoch": 7.885906040268456, + "flow_matching_loss": 0.014346472080796957, + "grad_norm": 0.12790009379386902, + "learning_rate": 2.9625929045422327e-07, + "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": 0.003967941668815911, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0014618574175983667, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.009397740941494703, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.014346472080796957, + "weighted_rkd_loss": 0.004698870470747352 + }, + { + "epoch": 7.902684563758389, + "flow_matching_loss": 0.01271637286990881, + "grad_norm": 0.16283373534679413, + "learning_rate": 2.7696108134781186e-07, + "loss": 0.0174, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.0039179531624540685, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0014930704259313644, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.009328976878896356, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.01271637286990881, + "weighted_rkd_loss": 0.004664488439448178 + }, + { + "epoch": 7.919463087248322, + "flow_matching_loss": 0.014059433806687593, + "grad_norm": 0.17524674534797668, + "learning_rate": 2.583069056597376e-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": 0.003773231990635395, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.001535862369928509, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.009082326386123895, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.014059433806687593, + "weighted_rkd_loss": 0.0045411631930619475 + }, + { + "epoch": 7.936241610738255, + "flow_matching_loss": 0.011765244835987686, + "grad_norm": 0.15201984345912933, + "learning_rate": 2.4029757938407116e-07, + "loss": 0.017, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.004287443915382027, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0018286535050719976, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.010403541289269925, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.011765244835987686, + "weighted_rkd_loss": 0.005201770644634962 + }, + { + "epoch": 7.953020134228188, + "flow_matching_loss": 0.013755477033555508, + "grad_norm": 0.19749130308628082, + "learning_rate": 2.229338903070871e-07, + "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": 0.0035599785624071956, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0014380234526470304, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.008557980647310615, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.013755477033555508, + "weighted_rkd_loss": 0.004278990323655307 + }, + { + "epoch": 7.969798657718121, + "flow_matching_loss": 0.014288169704377652, + "grad_norm": 0.17997831106185913, + "learning_rate": 2.0621659797279645e-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": 0.004082369967363775, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0016643157927319408, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.009829055797308684, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.014288169704377652, + "weighted_rkd_loss": 0.004914527898654342 + }, + { + "epoch": 7.9865771812080535, + "flow_matching_loss": 0.012982644187286497, + "grad_norm": 0.13272035121917725, + "learning_rate": 1.901464336497233e-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": 0.004022148088552058, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0015711100655607878, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.009615406207740308, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.012982644187286497, + "weighted_rkd_loss": 0.004807703103870154 + }, + { + "epoch": 8.003355704697986, + "flow_matching_loss": 0.012017120141535997, + "grad_norm": 0.1425294727087021, + "learning_rate": 1.7472410029891527e-07, + "loss": 0.0164, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.003690410265699029, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0013321076985448599, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.00871292818337679, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.012017120141535997, + "weighted_rkd_loss": 0.004356464091688395 + }, + { + "epoch": 8.020134228187919, + "flow_matching_loss": 0.011508079571649432, + "grad_norm": 0.1185975894331932, + "learning_rate": 1.5995027254319173e-07, + "loss": 0.0163, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.00405461045447737, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0014693920616991817, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.009578613005578518, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.011508079571649432, + "weighted_rkd_loss": 0.004789306502789259 + }, + { + "epoch": 8.036912751677852, + "flow_matching_loss": 0.011774006625637412, + "grad_norm": 0.12193300575017929, + "learning_rate": 1.4582559663764372e-07, + "loss": 0.0157, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.0032693485030904413, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0013653967354912311, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.00790409385226667, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.011774006625637412, + "weighted_rkd_loss": 0.003952046926133335 + }, + { + "epoch": 8.053691275167786, + "flow_matching_loss": 0.01376564335078001, + "grad_norm": 0.15403912961483002, + "learning_rate": 1.323506904413485e-07, + "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": 0.0037099570967257024, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0013822478882502764, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.008802162064239383, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.01376564335078001, + "weighted_rkd_loss": 0.004401081032119692 + }, + { + "epoch": 8.070469798657719, + "flow_matching_loss": 0.012571563199162483, + "grad_norm": 0.14847010374069214, + "learning_rate": 1.195261433903594e-07, + "loss": 0.0169, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.0035822264617308974, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0014063604292459786, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.008570813480764628, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.012571563199162483, + "weighted_rkd_loss": 0.004285406740382314 + }, + { + "epoch": 8.087248322147651, + "flow_matching_loss": 0.01303102783858776, + "grad_norm": 0.22479461133480072, + "learning_rate": 1.0735251647190969e-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": 0.004655194259248674, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0017578598228283227, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.011068248329684139, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.01303102783858776, + "weighted_rkd_loss": 0.005534124164842069 + }, + { + "epoch": 8.104026845637584, + "flow_matching_loss": 0.010825584968551993, + "grad_norm": 0.14044439792633057, + "learning_rate": 9.583034219987408e-08, + "loss": 0.0148, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.0032617375254631042, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0014044177252799273, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.007927892776206136, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.010825584968551993, + "weighted_rkd_loss": 0.003963946388103068 + }, + { + "epoch": 8.120805369127517, + "flow_matching_loss": 0.010878020897507668, + "grad_norm": 0.16086778044700623, + "learning_rate": 8.496012459148717e-08, + "loss": 0.0155, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.0038354229647666214, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0015930189460050316, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.009263864858075977, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.010878020897507668, + "weighted_rkd_loss": 0.004631932429037988 + }, + { + "epoch": 8.13758389261745, + "flow_matching_loss": 0.01375404316931963, + "grad_norm": 0.1751631796360016, + "learning_rate": 7.474233914527618e-08, + "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": 0.0038205950520932676, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0016478243051096797, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.00928901438601315, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.01375404316931963, + "weighted_rkd_loss": 0.004644507193006575 + }, + { + "epoch": 8.154362416107382, + "flow_matching_loss": 0.013494851160794496, + "grad_norm": 0.19929954409599304, + "learning_rate": 6.517743282028254e-08, + "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": 0.0034150957595556974, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0013123318669386208, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.008142523420974612, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.013494851160794496, + "weighted_rkd_loss": 0.004071261710487306 + }, + { + "epoch": 8.171140939597315, + "flow_matching_loss": 0.014098628470674158, + "grad_norm": 0.15721264481544495, + "learning_rate": 5.6265824016492604e-08, + "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": 0.0037533565191552043, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0015949857071973384, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.009101698687300087, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.014098628470674158, + "weighted_rkd_loss": 0.004550849343650043 + }, + { + "epoch": 8.187919463087248, + "flow_matching_loss": 0.013258166052401066, + "grad_norm": 0.12723150849342346, + "learning_rate": 4.800790255654719e-08, + "loss": 0.0173, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.0034274237463250757, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.001281585212564096, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.008136432617902756, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.013258166052401066, + "weighted_rkd_loss": 0.004068216308951378 + }, + { + "epoch": 8.204697986577182, + "flow_matching_loss": 0.01369173233397305, + "grad_norm": 0.21861012279987335, + "learning_rate": 4.040402966867862e-08, + "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": 0.0037710790522396563, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.001406640448840335, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.008948798477649688, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.01369173233397305, + "weighted_rkd_loss": 0.004474399238824844 + }, + { + "epoch": 8.221476510067115, + "flow_matching_loss": 0.011236160341650248, + "grad_norm": 0.1503625065088272, + "learning_rate": 3.3454537970925014e-08, + "loss": 0.0154, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.003518209815956652, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0013535513891838491, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.008389971125870943, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.011236160341650248, + "weighted_rkd_loss": 0.004194985562935472 + }, + { + "epoch": 8.238255033557047, + "flow_matching_loss": 0.013852568995207547, + "grad_norm": 0.16968147456645966, + "learning_rate": 2.715973145656192e-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": 0.0037199769169092177, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0014850097242742777, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.0089249636977911, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.013852568995207547, + "weighted_rkd_loss": 0.00446248184889555 + }, + { + "epoch": 8.25503355704698, + "flow_matching_loss": 0.011306985560804606, + "grad_norm": 0.14712770283222198, + "learning_rate": 2.151988548082129e-08, + "loss": 0.0158, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.003772911708801985, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0014175664517097175, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.008963389880955219, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.011306985560804606, + "weighted_rkd_loss": 0.0044816949404776095 + }, + { + "epoch": 8.271812080536913, + "flow_matching_loss": 0.012866842467337847, + "grad_norm": 0.14772823452949524, + "learning_rate": 1.6535246748837772e-08, + "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": 0.004458117461763322, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0017672449350357056, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.010683479905128478, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.012866842467337847, + "weighted_rkd_loss": 0.005341739952564239 + }, + { + "epoch": 8.288590604026846, + "flow_matching_loss": 0.013925638142973184, + "grad_norm": 0.17435146868228912, + "learning_rate": 1.2206033304855702e-08, + "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": 0.004066174430772662, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0015223037218675018, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.009654652560129762, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.013925638142973184, + "weighted_rkd_loss": 0.004827326280064881 + }, + { + "epoch": 8.305369127516778, + "flow_matching_loss": 0.013670520065352321, + "grad_norm": 0.22648420929908752, + "learning_rate": 8.532434522695031e-09, + "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": 0.004508587578311562, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0018952674232423305, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.010912442533299326, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.013670520065352321, + "weighted_rkd_loss": 0.005456221266649663 + }, + { + "epoch": 8.322147651006711, + "flow_matching_loss": 0.012735329382121564, + "grad_norm": 0.15040388703346252, + "learning_rate": 5.5146110974679805e-09, + "loss": 0.017, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.003590844781138003, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0014258236333262175, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.008607513178139925, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.012735329382121564, + "weighted_rkd_loss": 0.004303756589069963 + }, + { + "epoch": 8.338926174496644, + "flow_matching_loss": 0.011451990390196443, + "grad_norm": 0.1485985368490219, + "learning_rate": 3.1526950385496467e-09, + "loss": 0.0163, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.004111752426251769, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0014832181157544255, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.0097067229449749, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.011451990390196443, + "weighted_rkd_loss": 0.00485336147248745 + }, + { + "epoch": 8.355704697986576, + "flow_matching_loss": 0.011133158579468727, + "grad_norm": 0.15696369111537933, + "learning_rate": 1.4467896637959666e-09, + "loss": 0.0153, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.0035388516378588976, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0012879677873570473, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.008365671196952463, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.011133158579468727, + "weighted_rkd_loss": 0.0041828355984762314 + }, + { + "epoch": 8.37248322147651, + "flow_matching_loss": 0.013223122153431178, + "grad_norm": 0.17017419636249542, + "learning_rate": 3.9696959503898733e-10, + "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": 0.0038438115268945696, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.001413160125957802, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.00910078315064311, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.013223122153431178, + "weighted_rkd_loss": 0.004550391575321555 + }, + { + "epoch": 8.389261744966444, + "flow_matching_loss": 0.012224378576502205, + "grad_norm": 0.16432853043079376, + "learning_rate": 3.280754806156416e-12, + "loss": 0.0164, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.0034646759275346996, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0013615088420920074, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.008290860708802939, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.012224378576502205, + "weighted_rkd_loss": 0.004145430354401469 + }, + { + "epoch": 8.389261744966444, + "step": 5000, + "total_flos": 0.0, + "train_loss": 0.03723637925982475, + "train_runtime": 9180.8544, + "train_samples_per_second": 34.855, + "train_steps_per_second": 0.545 + } + ], + "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/train_5shot_seed0_fullactionhead_5k/training_args.bin b/FewShot/39tasks/5shot/FromRKD-Timewarp_FullactionHead/seed0/default/train_5shot_seed0_fullactionhead_5k/training_args.bin new file mode 100644 index 0000000000000000000000000000000000000000..9c72ac66853341dc75f7a66fe63230d9a9b71b85 --- /dev/null +++ b/FewShot/39tasks/5shot/FromRKD-Timewarp_FullactionHead/seed0/default/train_5shot_seed0_fullactionhead_5k/training_args.bin @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:87ff16bc6a1dabd0ec1cbda14c9c7677efea7066e6b379642c32d9370fc6a332 +size 6033 diff --git a/FewShot/39tasks/5shot/FromRKD-Timewarp_FullactionHead/seed1/default/train_5shot_seed1_fullactionhead_5k/checkpoint-5000/config.json b/FewShot/39tasks/5shot/FromRKD-Timewarp_FullactionHead/seed1/default/train_5shot_seed1_fullactionhead_5k/checkpoint-5000/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-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-Timewarp_FullactionHead/seed1/default/train_5shot_seed1_fullactionhead_5k/checkpoint-5000/experiment_cfg/metadata.json b/FewShot/39tasks/5shot/FromRKD-Timewarp_FullactionHead/seed1/default/train_5shot_seed1_fullactionhead_5k/checkpoint-5000/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-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": [ + 2.5719731235192254, + -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-5000/model-00001-of-00002.safetensors b/FewShot/39tasks/5shot/FromRKD-Timewarp_FullactionHead/seed1/default/train_5shot_seed1_fullactionhead_5k/checkpoint-5000/model-00001-of-00002.safetensors new file mode 100644 index 0000000000000000000000000000000000000000..46d8e7b66bb949d86548e5d76fb38ce09b265acf --- /dev/null +++ b/FewShot/39tasks/5shot/FromRKD-Timewarp_FullactionHead/seed1/default/train_5shot_seed1_fullactionhead_5k/checkpoint-5000/model-00001-of-00002.safetensors @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:d3cb77271d1a9fce833f206b6913da1e083de735183e622ded352d7b12681fb1 +size 4999367032 diff --git a/FewShot/39tasks/5shot/FromRKD-Timewarp_FullactionHead/seed1/default/train_5shot_seed1_fullactionhead_5k/checkpoint-5000/model-00002-of-00002.safetensors b/FewShot/39tasks/5shot/FromRKD-Timewarp_FullactionHead/seed1/default/train_5shot_seed1_fullactionhead_5k/checkpoint-5000/model-00002-of-00002.safetensors new file mode 100644 index 0000000000000000000000000000000000000000..3c984996137bf66d92f145056d386104dba4a503 --- /dev/null +++ b/FewShot/39tasks/5shot/FromRKD-Timewarp_FullactionHead/seed1/default/train_5shot_seed1_fullactionhead_5k/checkpoint-5000/model-00002-of-00002.safetensors @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:8e1d911f2e4331b03e74d36773631a8e6ac5b184267257c9c86e233b51cf80f3 +size 2664540348 diff --git a/FewShot/39tasks/5shot/FromRKD-Timewarp_FullactionHead/seed1/default/train_5shot_seed1_fullactionhead_5k/checkpoint-5000/model.safetensors.index.json b/FewShot/39tasks/5shot/FromRKD-Timewarp_FullactionHead/seed1/default/train_5shot_seed1_fullactionhead_5k/checkpoint-5000/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-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-Timewarp_FullactionHead/seed1/default/train_5shot_seed1_fullactionhead_5k/checkpoint-5000/optimizer.pt b/FewShot/39tasks/5shot/FromRKD-Timewarp_FullactionHead/seed1/default/train_5shot_seed1_fullactionhead_5k/checkpoint-5000/optimizer.pt new file mode 100644 index 0000000000000000000000000000000000000000..b06061e2b230fb907b0b782ffebc4013eb9025a5 --- /dev/null +++ b/FewShot/39tasks/5shot/FromRKD-Timewarp_FullactionHead/seed1/default/train_5shot_seed1_fullactionhead_5k/checkpoint-5000/optimizer.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:82811a21c277b2bc53ee264d227d38451f2c82a8ed18435eacd8a4ae5beade78 +size 8563326263 diff --git a/FewShot/39tasks/5shot/FromRKD-Timewarp_FullactionHead/seed1/default/train_5shot_seed1_fullactionhead_5k/checkpoint-5000/rng_state.pth b/FewShot/39tasks/5shot/FromRKD-Timewarp_FullactionHead/seed1/default/train_5shot_seed1_fullactionhead_5k/checkpoint-5000/rng_state.pth new file mode 100644 index 0000000000000000000000000000000000000000..051c6975a6fc4e94fd851ad5eea3bc7a6763e8de --- /dev/null +++ b/FewShot/39tasks/5shot/FromRKD-Timewarp_FullactionHead/seed1/default/train_5shot_seed1_fullactionhead_5k/checkpoint-5000/rng_state.pth @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:54f950b8bb9b0c32184ffeb79794242a6b8e05bbd5987343fab0e5c1c77ad754 +size 14645 diff --git a/FewShot/39tasks/5shot/FromRKD-Timewarp_FullactionHead/seed1/default/train_5shot_seed1_fullactionhead_5k/checkpoint-5000/scheduler.pt b/FewShot/39tasks/5shot/FromRKD-Timewarp_FullactionHead/seed1/default/train_5shot_seed1_fullactionhead_5k/checkpoint-5000/scheduler.pt new file mode 100644 index 0000000000000000000000000000000000000000..8cb18c42a19c524a920dd1ec232118cfa1c2aa5a --- /dev/null +++ b/FewShot/39tasks/5shot/FromRKD-Timewarp_FullactionHead/seed1/default/train_5shot_seed1_fullactionhead_5k/checkpoint-5000/scheduler.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:39f93e93eb5653777630c590a20e98e9373f48f53da024592383b624fc2dcbde +size 1465 diff --git a/FewShot/39tasks/5shot/FromRKD-Timewarp_FullactionHead/seed1/default/train_5shot_seed1_fullactionhead_5k/checkpoint-5000/trainer_state.json b/FewShot/39tasks/5shot/FromRKD-Timewarp_FullactionHead/seed1/default/train_5shot_seed1_fullactionhead_5k/checkpoint-5000/trainer_state.json new file mode 100644 index 0000000000000000000000000000000000000000..262a17e9dacc43c7583a1d07222860fd75c3826b --- /dev/null +++ b/FewShot/39tasks/5shot/FromRKD-Timewarp_FullactionHead/seed1/default/train_5shot_seed1_fullactionhead_5k/checkpoint-5000/trainer_state.json @@ -0,0 +1,23034 @@ +{ + "best_global_step": null, + "best_metric": null, + "best_model_checkpoint": null, + "epoch": 8.375209380234505, + "eval_steps": 500, + "global_step": 5000, + "is_hyper_param_search": false, + "is_local_process_zero": true, + "is_world_process_zero": true, + "log_history": [ + { + "epoch": 0.01675041876046901, + "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 + }, + { + "epoch": 5.041876046901173, + "flow_matching_loss": 0.015169484168291092, + "grad_norm": 0.21513906121253967, + "learning_rate": 1.1231228239056467e-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": 0.005936745693907142, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0022759507060982286, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01414944203570485, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.015169484168291092, + "weighted_rkd_loss": 0.007074721017852425 + }, + { + "epoch": 5.058626465661642, + "flow_matching_loss": 0.016462551802396773, + "grad_norm": 0.1520560383796692, + "learning_rate": 1.113528559434845e-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": 0.0047940189950168135, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0019913177471607923, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.011579355597496033, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.016462551802396773, + "weighted_rkd_loss": 0.0057896777987480165 + }, + { + "epoch": 5.075376884422111, + "flow_matching_loss": 0.017612619884312154, + "grad_norm": 0.17633816599845886, + "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": 0.005207244050689041, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0022622236749157308, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.012676711846143007, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.017612619884312154, + "weighted_rkd_loss": 0.006338355923071503 + }, + { + "epoch": 5.09212730318258, + "flow_matching_loss": 0.015209727641195058, + "grad_norm": 0.1599225252866745, + "learning_rate": 1.0943911659763612e-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": 0.005452072992920875, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0021627967827953396, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.013066942524164914, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.015209727641195058, + "weighted_rkd_loss": 0.006533471262082457 + }, + { + "epoch": 5.108877721943049, + "flow_matching_loss": 0.014849191531538963, + "grad_norm": 0.15893946588039398, + "learning_rate": 1.084848874120266e-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": 0.005263407132588327, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0025313270860351624, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.013058141246438027, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.014849191531538963, + "weighted_rkd_loss": 0.006529070623219013 + }, + { + "epoch": 5.125628140703517, + "flow_matching_loss": 0.01759036611765623, + "grad_norm": 0.22340935468673706, + "learning_rate": 1.0753247423193196e-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": 0.005642558354884386, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.002568938955664635, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.013854055665433407, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.01759036611765623, + "weighted_rkd_loss": 0.006927027832716703 + }, + { + "epoch": 5.142378559463987, + "flow_matching_loss": 0.015588127821683884, + "grad_norm": 0.1604069173336029, + "learning_rate": 1.0658191871899012e-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": 0.005148043041117489, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.002302269684150815, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.012598355766385794, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.015588127821683884, + "weighted_rkd_loss": 0.006299177883192897 + }, + { + "epoch": 5.159128978224456, + "flow_matching_loss": 0.01548425005748868, + "grad_norm": 0.17967723309993744, + "learning_rate": 1.0563326245357852e-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": 0.004702624608762562, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0019607308669947086, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.011365980049595237, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.01548425005748868, + "weighted_rkd_loss": 0.0056829900247976186 + }, + { + "epoch": 5.175879396984925, + "flow_matching_loss": 0.014517775271087885, + "grad_norm": 0.20446251332759857, + "learning_rate": 1.0468654693299557e-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": 0.00531402169726789, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0026462510228157043, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.013274294324219227, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.014517775271087885, + "weighted_rkd_loss": 0.006637147162109613 + }, + { + "epoch": 5.192629815745394, + "flow_matching_loss": 0.01355633307248354, + "grad_norm": 0.20802079141139984, + "learning_rate": 1.0374181356964501e-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": 0.006404514703899622, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0025749327382072807, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.015383962262421846, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.01355633307248354, + "weighted_rkd_loss": 0.007691981131210923 + }, + { + "epoch": 5.209380234505863, + "flow_matching_loss": 0.014503863453865052, + "grad_norm": 0.18346713483333588, + "learning_rate": 1.0279910368922465e-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": 0.005380265135318041, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0024684223462827504, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.0132289526052773, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.014503863453865052, + "weighted_rkd_loss": 0.00661447630263865 + }, + { + "epoch": 5.226130653266332, + "flow_matching_loss": 0.013800499308854342, + "grad_norm": 0.1441630870103836, + "learning_rate": 1.018584585289186e-05, + "loss": 0.0197, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.004836055263876915, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0021854664431884886, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.011857577040791512, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.013800499308854342, + "weighted_rkd_loss": 0.005928788520395756 + }, + { + "epoch": 5.242881072026801, + "flow_matching_loss": 0.015614958759397269, + "grad_norm": 0.22036311030387878, + "learning_rate": 1.0091991923559348e-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": 0.00548774020280689, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0024738425039686263, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.013449322991073131, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.015614958759397269, + "weighted_rkd_loss": 0.006724661495536566 + }, + { + "epoch": 5.259631490787269, + "flow_matching_loss": 0.014919089665636421, + "grad_norm": 0.1609790325164795, + "learning_rate": 9.99835268639984e-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": 0.005167843447998166, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0021391444955952466, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.012474831566214562, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.014919089665636421, + "weighted_rkd_loss": 0.006237415783107281 + }, + { + "epoch": 5.276381909547739, + "flow_matching_loss": 0.013936952222138643, + "grad_norm": 0.16832859814167023, + "learning_rate": 9.904932237496913e-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": 0.0049352750414982435, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.002076266333460808, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.011946816369891167, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.013936952222138643, + "weighted_rkd_loss": 0.0059734081849455835 + }, + { + "epoch": 5.293132328308207, + "flow_matching_loss": 0.012869645096361636, + "grad_norm": 0.17330895364284515, + "learning_rate": 9.811734663363659e-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": 0.005114914570003748, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0023217051057145, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.012551534455269575, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.012869645096361636, + "weighted_rkd_loss": 0.0062757672276347876 + }, + { + "epoch": 5.309882747068677, + "flow_matching_loss": 0.01596754267811775, + "grad_norm": 0.20717377960681915, + "learning_rate": 9.718764040763875e-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": 0.0068621003534644846, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.002769121224991977, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.016493321862071753, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.01596754267811775, + "weighted_rkd_loss": 0.008246660931035876 + }, + { + "epoch": 5.326633165829146, + "flow_matching_loss": 0.013056589663028717, + "grad_norm": 0.18150459229946136, + "learning_rate": 9.626024436533777e-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": 0.005945549625903368, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.002427968825213611, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.014319068286567926, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.013056589663028717, + "weighted_rkd_loss": 0.007159534143283963 + }, + { + "epoch": 5.343383584589615, + "flow_matching_loss": 0.01568213803693652, + "grad_norm": 0.1437157690525055, + "learning_rate": 9.533519907404095e-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": 0.0047857168596237894, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.002092471101786941, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.011663904972374439, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.01568213803693652, + "weighted_rkd_loss": 0.0058319524861872194 + }, + { + "epoch": 5.360134003350084, + "flow_matching_loss": 0.017202568333595993, + "grad_norm": 0.1709163784980774, + "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": 0.00488214164506644, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0021265184856019913, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.011890801694244146, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.017202568333595993, + "weighted_rkd_loss": 0.005945400847122073 + }, + { + "epoch": 5.376884422110553, + "flow_matching_loss": 0.01590415444225073, + "grad_norm": 0.1625984013080597, + "learning_rate": 9.349232249777124e-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": 0.005498569365590811, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0023033414385281504, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01330048032104969, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.01590415444225073, + "weighted_rkd_loss": 0.006650240160524845 + }, + { + "epoch": 5.393634840871022, + "flow_matching_loss": 0.0181461731903255, + "grad_norm": 0.19519178569316864, + "learning_rate": 9.257457182618985e-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": 0.005043261009268463, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.002520814340095967, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.012607336603105068, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.0181461731903255, + "weighted_rkd_loss": 0.006303668301552534 + }, + { + "epoch": 5.410385259631491, + "flow_matching_loss": 0.014780774526298047, + "grad_norm": 0.18064932525157928, + "learning_rate": 9.165933312886944e-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": 0.006047557084821164, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0027511882479302587, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.014846302475780249, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.014780774526298047, + "weighted_rkd_loss": 0.0074231512378901245 + }, + { + "epoch": 5.42713567839196, + "flow_matching_loss": 0.013527859747409821, + "grad_norm": 0.15953907370567322, + "learning_rate": 9.07466464413155e-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": 0.004121492733247578, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0020615094457753004, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.010304494947195052, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.013527859747409821, + "weighted_rkd_loss": 0.005152247473597526 + }, + { + "epoch": 5.443886097152429, + "flow_matching_loss": 0.016063749976456165, + "grad_norm": 0.24305878579616547, + "learning_rate": 8.983655168740048e-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": 0.0049547473667189475, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0020637805806472897, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01197327538393438, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.016063749976456165, + "weighted_rkd_loss": 0.00598663769196719 + }, + { + "epoch": 5.460636515912898, + "flow_matching_loss": 0.013650599587708711, + "grad_norm": 0.1641220897436142, + "learning_rate": 8.892908867761718e-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": 0.004460714524611831, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.00191993199987337, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.010841361153870821, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.013650599587708711, + "weighted_rkd_loss": 0.005420680576935411 + }, + { + "epoch": 5.477386934673367, + "flow_matching_loss": 0.016264970973134042, + "grad_norm": 0.19726066291332245, + "learning_rate": 8.802429710733745e-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": 0.004725346667692065, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.002034914761316031, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.011485608201473952, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.016264970973134042, + "weighted_rkd_loss": 0.005742804100736976 + }, + { + "epoch": 5.494137353433836, + "flow_matching_loss": 0.014542607683688402, + "grad_norm": 0.14502941071987152, + "learning_rate": 8.71222165550757e-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": 0.005226415698416531, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0020495078642852604, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.012502339389175176, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.014542607683688402, + "weighted_rkd_loss": 0.006251169694587588 + }, + { + "epoch": 5.510887772194305, + "flow_matching_loss": 0.015074159018695354, + "grad_norm": 0.1566695272922516, + "learning_rate": 8.622288648075748e-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": 0.004941701982170343, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.00232604939956218, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.012209453340619802, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.015074159018695354, + "weighted_rkd_loss": 0.006104726670309901 + }, + { + "epoch": 5.527638190954773, + "flow_matching_loss": 0.01538830967620015, + "grad_norm": 0.1672288030385971, + "learning_rate": 8.532634622399391e-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": 0.0053394132293760775, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0020577054703608154, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.012736531719565392, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.01538830967620015, + "weighted_rkd_loss": 0.006368265859782696 + }, + { + "epoch": 5.544388609715243, + "flow_matching_loss": 0.012408527079969645, + "grad_norm": 0.13449180126190186, + "learning_rate": 8.44326350023601e-06, + "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": 0.004540071333758533, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0017385528131853788, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.010818695602938532, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.012408527079969645, + "weighted_rkd_loss": 0.005409347801469266 + }, + { + "epoch": 5.561139028475711, + "flow_matching_loss": 0.014972069207578898, + "grad_norm": 0.1624370515346527, + "learning_rate": 8.354179190968002e-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": 0.005195741541683674, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.002315680112224072, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.012707163114100695, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.014972069207578898, + "weighted_rkd_loss": 0.006353581557050347 + }, + { + "epoch": 5.577889447236181, + "flow_matching_loss": 0.017658597324043512, + "grad_norm": 0.1527981162071228, + "learning_rate": 8.265385591431661e-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": 0.005545643903315067, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0025704527739435433, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.013661740627139807, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.017658597324043512, + "weighted_rkd_loss": 0.006830870313569904 + }, + { + "epoch": 5.594639865996649, + "flow_matching_loss": 0.013222060818225145, + "grad_norm": 0.1373794972896576, + "learning_rate": 8.176886585746671e-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": 0.005418227729387581, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0022093447158113124, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.013045800291001797, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.013222060818225145, + "weighted_rkd_loss": 0.0065229001455008985 + }, + { + "epoch": 5.611390284757119, + "flow_matching_loss": 0.012797685340046883, + "grad_norm": 0.16945552825927734, + "learning_rate": 8.088686045146247e-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": 0.004318593186326325, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0023255046922713517, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.010962691204622389, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.012797685340046883, + "weighted_rkd_loss": 0.005481345602311194 + }, + { + "epoch": 5.628140703517588, + "flow_matching_loss": 0.013005319610238075, + "grad_norm": 0.1570693403482437, + "learning_rate": 8.00078782780775e-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": 0.004446164006367326, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0022311551845632495, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.011123483255505561, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.013005319610238075, + "weighted_rkd_loss": 0.005561741627752781 + }, + { + "epoch": 5.644891122278057, + "flow_matching_loss": 0.01654042676091194, + "grad_norm": 0.15214891731739044, + "learning_rate": 7.913195778683979e-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": 0.0048288710881024596, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.00192901233676821, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.011586754489690066, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.01654042676091194, + "weighted_rkd_loss": 0.005793377244845033 + }, + { + "epoch": 5.661641541038526, + "flow_matching_loss": 0.014083790592849255, + "grad_norm": 0.15469446778297424, + "learning_rate": 7.825913729334908e-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": 0.00407305343542248, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0017037070298101752, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.009849813859909773, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.014083790592849255, + "weighted_rkd_loss": 0.004924906929954886 + }, + { + "epoch": 5.678391959798995, + "flow_matching_loss": 0.016759827733039856, + "grad_norm": 0.20145344734191895, + "learning_rate": 7.73894549776013e-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": 0.003940209886059165, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0016941382200457156, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.009574558027088642, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.016759827733039856, + "weighted_rkd_loss": 0.004787279013544321 + }, + { + "epoch": 5.695142378559464, + "flow_matching_loss": 0.013244044221937656, + "grad_norm": 0.2122158706188202, + "learning_rate": 7.65229488823184e-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": 0.005202788161113858, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0020280405762605367, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.012433616816997528, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.013244044221937656, + "weighted_rkd_loss": 0.006216808408498764 + }, + { + "epoch": 5.711892797319933, + "flow_matching_loss": 0.01592979095876217, + "grad_norm": 0.18316158652305603, + "learning_rate": 7.565965691128396e-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": 0.005638341279700398, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.002074102347251028, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.013350784964859485, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.01592979095876217, + "weighted_rkd_loss": 0.0066753924824297425 + }, + { + "epoch": 5.728643216080402, + "flow_matching_loss": 0.013631716649979354, + "grad_norm": 0.16366751492023468, + "learning_rate": 7.47996168276856e-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": 0.0039018418407067656, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0014133890508674085, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.009217072697356343, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.013631716649979354, + "weighted_rkd_loss": 0.004608536348678171 + }, + { + "epoch": 5.745393634840871, + "flow_matching_loss": 0.014068759791553021, + "grad_norm": 0.14519792795181274, + "learning_rate": 7.394286625246258e-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": 0.005458051455207169, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.002536578814033419, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.013452681805938482, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.014068759791553021, + "weighted_rkd_loss": 0.006726340902969241 + }, + { + "epoch": 5.76214405360134, + "flow_matching_loss": 0.013327827304601669, + "grad_norm": 0.15125919878482819, + "learning_rate": 7.308944266266063e-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": 0.004904886498115957, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.002017999248346314, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.011827772436663508, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.013327827304601669, + "weighted_rkd_loss": 0.005913886218331754 + }, + { + "epoch": 5.778894472361809, + "flow_matching_loss": 0.01448378199711442, + "grad_norm": 0.16186784207820892, + "learning_rate": 7.223938338979233e-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": 0.004507014737464487, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0019386717118322848, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01095270118676126, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.01448378199711442, + "weighted_rkd_loss": 0.00547635059338063 + }, + { + "epoch": 5.795644891122278, + "flow_matching_loss": 0.015600308123975993, + "grad_norm": 0.1659005880355835, + "learning_rate": 7.139272561820397e-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": 0.004978228081017733, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.00196147394599393, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.011917930282652377, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.015600308123975993, + "weighted_rkd_loss": 0.005958965141326189 + }, + { + "epoch": 5.812395309882747, + "flow_matching_loss": 0.01297576930373907, + "grad_norm": 0.16415287554264069, + "learning_rate": 7.054950638344936e-06, + "loss": 0.0179, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.004142170492559672, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.001531300344504416, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.009815641259774566, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.01297576930373907, + "weighted_rkd_loss": 0.004907820629887283 + }, + { + "epoch": 5.8291457286432165, + "flow_matching_loss": 0.013271164055913687, + "grad_norm": 0.16550561785697937, + "learning_rate": 6.970976257066941e-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": 0.0046145610976964235, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0018462048145011067, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.011075327079743147, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.013271164055913687, + "weighted_rkd_loss": 0.005537663539871574 + }, + { + "epoch": 5.8458961474036855, + "flow_matching_loss": 0.015127340983599424, + "grad_norm": 0.15099075436592102, + "learning_rate": 6.887353091297896e-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": 0.004318542499095202, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0018304343800991773, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01046751942485571, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.015127340983599424, + "weighted_rkd_loss": 0.005233759712427855 + }, + { + "epoch": 5.8626465661641545, + "flow_matching_loss": 0.01355399824678898, + "grad_norm": 0.14437517523765564, + "learning_rate": 6.804084798985965e-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": 0.004264809051528573, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0019032558193430305, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01043287394568324, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.01355399824678898, + "weighted_rkd_loss": 0.00521643697284162 + }, + { + "epoch": 5.8793969849246235, + "flow_matching_loss": 0.014860629010945559, + "grad_norm": 0.1718394011259079, + "learning_rate": 6.721175022556009e-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": 0.00439409245736897, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0017511130310595036, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.010539297945797443, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.014860629010945559, + "weighted_rkd_loss": 0.005269648972898722 + }, + { + "epoch": 5.8961474036850925, + "flow_matching_loss": 0.01369728147983551, + "grad_norm": 0.20387037098407745, + "learning_rate": 6.638627388750245e-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": 0.004711744654923677, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0019726135593373328, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.0113961027469486, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.01369728147983551, + "weighted_rkd_loss": 0.0056980513734743 + }, + { + "epoch": 5.9128978224455615, + "flow_matching_loss": 0.01615075683221221, + "grad_norm": 0.1426304429769516, + "learning_rate": 6.556445508469583e-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": 0.004828235320746899, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.001961126539390534, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.011617597192525864, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.01615075683221221, + "weighted_rkd_loss": 0.005808798596262932 + }, + { + "epoch": 5.9296482412060305, + "flow_matching_loss": 0.013974253181368112, + "grad_norm": 0.14448580145835876, + "learning_rate": 6.474632976615708e-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": 0.004280661465600133, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0018652994302101433, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.010426622349768877, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.013974253181368112, + "weighted_rkd_loss": 0.0052133111748844385 + }, + { + "epoch": 5.9463986599664995, + "flow_matching_loss": 0.013339675264433026, + "grad_norm": 0.14109483361244202, + "learning_rate": 6.393193371933791e-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": 0.004555360320955515, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0015687765087932348, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.010679497197270394, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.013339675264433026, + "weighted_rkd_loss": 0.005339748598635197 + }, + { + "epoch": 5.9631490787269685, + "flow_matching_loss": 0.012223204039037228, + "grad_norm": 0.15760351717472076, + "learning_rate": 6.312130256855977e-06, + "loss": 0.0172, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.004092269181273878, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0018160081468522548, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.010000546555966138, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.012223204039037228, + "weighted_rkd_loss": 0.005000273277983069 + }, + { + "epoch": 5.9798994974874375, + "flow_matching_loss": 0.013196866121143102, + "grad_norm": 0.13676081597805023, + "learning_rate": 6.23144717734552e-06, + "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": 0.0035694713238626718, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0014613688690587878, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.008600311633199454, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.013196866121143102, + "weighted_rkd_loss": 0.004300155816599727 + }, + { + "epoch": 5.9966499162479066, + "flow_matching_loss": 0.014821585360914468, + "grad_norm": 0.23386721312999725, + "learning_rate": 6.151147662741695e-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": 0.005108627933077514, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0021549923927523195, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01237224810756743, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.014821585360914468, + "weighted_rkd_loss": 0.006186124053783715 + }, + { + "epoch": 6.013400335008376, + "flow_matching_loss": 0.015139726269990206, + "grad_norm": 0.19289110600948334, + "learning_rate": 6.071235225605415e-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": 0.004597576847299933, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0017556668259203433, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.010950820334255696, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.015139726269990206, + "weighted_rkd_loss": 0.005475410167127848 + }, + { + "epoch": 6.030150753768845, + "flow_matching_loss": 0.013879562634974718, + "grad_norm": 0.21247315406799316, + "learning_rate": 5.991713361565546e-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": 0.004605715093202889, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0019215092528611422, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01113293943926692, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.013879562634974718, + "weighted_rkd_loss": 0.00556646971963346 + }, + { + "epoch": 6.046901172529314, + "flow_matching_loss": 0.011976712848991156, + "grad_norm": 0.20955568552017212, + "learning_rate": 5.912585549166051e-06, + "loss": 0.0172, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.004337595240212977, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0017820636508986353, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.010457254108041525, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.011976712848991156, + "weighted_rkd_loss": 0.005228627054020762 + }, + { + "epoch": 6.063651591289783, + "flow_matching_loss": 0.014497782150283457, + "grad_norm": 0.14614950120449066, + "learning_rate": 5.833855249713777e-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": 0.004382647341117263, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0017091153538785874, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.010474410047754646, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.014497782150283457, + "weighted_rkd_loss": 0.005237205023877323 + }, + { + "epoch": 6.080402010050252, + "flow_matching_loss": 0.014638572558760643, + "grad_norm": 0.17804445326328278, + "learning_rate": 5.7555259071270645e-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": 0.004319529328495264, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.002127594541525468, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.010766653111204505, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.014638572558760643, + "weighted_rkd_loss": 0.0053833265556022525 + }, + { + "epoch": 6.097152428810721, + "flow_matching_loss": 0.012663161847740412, + "grad_norm": 0.14908669888973236, + "learning_rate": 5.677600947785117e-06, + "loss": 0.0171, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.0037628420628607272, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0013517275685444475, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.00887741162441671, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.012663161847740412, + "weighted_rkd_loss": 0.004438705812208355 + }, + { + "epoch": 6.11390284757119, + "flow_matching_loss": 0.014881637273356318, + "grad_norm": 0.2055513858795166, + "learning_rate": 5.600083780378099e-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": 0.004984625265933573, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.002195600490085781, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.012164850998669863, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.014881637273356318, + "weighted_rkd_loss": 0.006082425499334932 + }, + { + "epoch": 6.130653266331659, + "flow_matching_loss": 0.014626818615943194, + "grad_norm": 0.18061131238937378, + "learning_rate": 5.522977795758038e-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": 0.004626819328404963, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.001866737351519987, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.011120375851169229, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.014626818615943194, + "weighted_rkd_loss": 0.0055601879255846145 + }, + { + "epoch": 6.147403685092128, + "flow_matching_loss": 0.013677647057920694, + "grad_norm": 0.1852584183216095, + "learning_rate": 5.4462863667904805e-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": 0.00379023642744869, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0017943391227163374, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.0093748118262738, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.013677647057920694, + "weighted_rkd_loss": 0.0046874059131369 + }, + { + "epoch": 6.164154103852597, + "flow_matching_loss": 0.013752449536696076, + "grad_norm": 0.15452587604522705, + "learning_rate": 5.370012848206983e-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": 0.003933401824906468, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0015195835614576936, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.009386387234553695, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.013752449536696076, + "weighted_rkd_loss": 0.004693193617276848 + }, + { + "epoch": 6.180904522613066, + "flow_matching_loss": 0.013424647925421595, + "grad_norm": 0.16993391513824463, + "learning_rate": 5.29416057645834e-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": 0.004384127364028245, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0018952976912260055, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01066355244256556, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.013424647925421595, + "weighted_rkd_loss": 0.00533177622128278 + }, + { + "epoch": 6.197654941373535, + "flow_matching_loss": 0.014398802910000086, + "grad_norm": 0.15566317737102509, + "learning_rate": 5.218732869568635e-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": 0.005112427612766623, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0022814268944785, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01250628214329481, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.014398802910000086, + "weighted_rkd_loss": 0.006253141071647405 + }, + { + "epoch": 6.214405360134004, + "flow_matching_loss": 0.013579555507749318, + "grad_norm": 0.15493236482143402, + "learning_rate": 5.14373302699013e-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": 0.004654577840119601, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0017654628201853484, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.011074618343263865, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.013579555507749318, + "weighted_rkd_loss": 0.005537309171631932 + }, + { + "epoch": 6.231155778894473, + "flow_matching_loss": 0.012991676107048988, + "grad_norm": 0.14714433252811432, + "learning_rate": 5.069164329458894e-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": 0.004408569377847016, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0017401628778316081, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01055730152875185, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.012991676107048988, + "weighted_rkd_loss": 0.005278650764375925 + }, + { + "epoch": 6.247906197654942, + "flow_matching_loss": 0.014295129012316466, + "grad_norm": 0.17364512383937836, + "learning_rate": 4.995030038851342e-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": 0.004136458213906735, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0018542096950113773, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.010127126052975655, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.014295129012316466, + "weighted_rkd_loss": 0.005063563026487828 + }, + { + "epoch": 6.264656616415411, + "flow_matching_loss": 0.01341222645714879, + "grad_norm": 0.20091146230697632, + "learning_rate": 4.9213333980415e-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": 0.0047332413727417585, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0018039730959571898, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.011270455876365304, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.01341222645714879, + "weighted_rkd_loss": 0.005635227938182652 + }, + { + "epoch": 6.28140703517588, + "flow_matching_loss": 0.014535247907042503, + "grad_norm": 0.14015261828899384, + "learning_rate": 4.848077630759175e-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": 0.004209902766160667, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.00176135414512828, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01018115975894034, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.014535247907042503, + "weighted_rkd_loss": 0.00509057987947017 + }, + { + "epoch": 6.298157453936349, + "flow_matching_loss": 0.013406949071213602, + "grad_norm": 0.16716063022613525, + "learning_rate": 4.775265941448951e-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": 0.003995424625463784, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0015201234142296015, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.00951097272336483, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.013406949071213602, + "weighted_rkd_loss": 0.004755486361682415 + }, + { + "epoch": 6.314907872696818, + "flow_matching_loss": 0.013139669876545667, + "grad_norm": 0.14437201619148254, + "learning_rate": 4.702901515129989e-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": 0.0044357549166306855, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0017475514672696591, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.010619061347097159, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.013139669876545667, + "weighted_rkd_loss": 0.0053095306735485796 + }, + { + "epoch": 6.331658291457287, + "flow_matching_loss": 0.013105375785380602, + "grad_norm": 0.13337287306785583, + "learning_rate": 4.630987517256734e-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": 0.00446128414478153, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.00155455416534096, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.010477122571319341, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.013105375785380602, + "weighted_rkd_loss": 0.0052385612856596705 + }, + { + "epoch": 6.348408710217756, + "flow_matching_loss": 0.01332718194462359, + "grad_norm": 0.15027029812335968, + "learning_rate": 4.559527093580412e-06, + "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": 0.004570346092805266, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.002078494516899809, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.011219186754897238, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.01332718194462359, + "weighted_rkd_loss": 0.005609593377448619 + }, + { + "epoch": 6.365159128978225, + "flow_matching_loss": 0.0158365941606462, + "grad_norm": 0.15022490918636322, + "learning_rate": 4.488523370011456e-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": 0.004662989941425622, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0019907014677301047, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.011316681420430541, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.0158365941606462, + "weighted_rkd_loss": 0.005658340710215271 + }, + { + "epoch": 6.381909547738694, + "flow_matching_loss": 0.012740066088736058, + "grad_norm": 0.21718840301036835, + "learning_rate": 4.417979452482766e-06, + "loss": 0.0179, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.00425658579915762, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0017564637295436115, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01026963540352881, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.012740066088736058, + "weighted_rkd_loss": 0.005134817701764405 + }, + { + "epoch": 6.398659966499163, + "flow_matching_loss": 0.012215562909841538, + "grad_norm": 0.1450190246105194, + "learning_rate": 4.347898426813811e-06, + "loss": 0.0179, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.004763328027911485, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0018555120564997195, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.011382168158888817, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.012215562909841538, + "weighted_rkd_loss": 0.005691084079444408 + }, + { + "epoch": 6.415410385259632, + "flow_matching_loss": 0.01273781326599419, + "grad_norm": 0.14332900941371918, + "learning_rate": 4.278283358575694e-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": 0.0041654838249087335, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0016894752159714698, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01002044277265668, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.01273781326599419, + "weighted_rkd_loss": 0.00501022138632834 + }, + { + "epoch": 6.432160804020101, + "flow_matching_loss": 0.013954770937561988, + "grad_norm": 0.16812093555927277, + "learning_rate": 4.2091372929570145e-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": 0.003278344660066068, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0015723046730272473, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.008128994004800915, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.013954770937561988, + "weighted_rkd_loss": 0.004064497002400458 + }, + { + "epoch": 6.44891122278057, + "flow_matching_loss": 0.015501107182353735, + "grad_norm": 0.16985221207141876, + "learning_rate": 4.1404632546306934e-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": 0.004027113923802972, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0016444221953861415, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.009698650008067489, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.015501107182353735, + "weighted_rkd_loss": 0.0048493250040337445 + }, + { + "epoch": 6.465661641541039, + "flow_matching_loss": 0.014016295969486236, + "grad_norm": 0.16254419088363647, + "learning_rate": 4.0722642476216286e-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": 0.0042627822840586305, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0018491683877073228, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.010374732967466115, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.014016295969486236, + "weighted_rkd_loss": 0.005187366483733058 + }, + { + "epoch": 6.482412060301508, + "flow_matching_loss": 0.013069759542122483, + "grad_norm": 0.1539461463689804, + "learning_rate": 4.00454325517533e-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": 0.004568273038603366, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.002030610025394708, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.011167156137526035, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.013069759542122483, + "weighted_rkd_loss": 0.0055835780687630175 + }, + { + "epoch": 6.499162479061977, + "flow_matching_loss": 0.013110130187124014, + "grad_norm": 0.1754523068666458, + "learning_rate": 3.937303239627393e-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": 0.004449385032057762, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0018155326717533172, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.010714302537962794, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.013110130187124014, + "weighted_rkd_loss": 0.005357151268981397 + }, + { + "epoch": 6.515912897822446, + "flow_matching_loss": 0.014289885107427836, + "grad_norm": 0.13485778868198395, + "learning_rate": 3.870547142273918e-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": 0.0033146835397928954, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0012253299821168185, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.007854697108268738, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.014289885107427836, + "weighted_rkd_loss": 0.003927348554134369 + }, + { + "epoch": 6.532663316582915, + "flow_matching_loss": 0.013430317863821983, + "grad_norm": 0.13589271903038025, + "learning_rate": 3.8042778832428736e-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": 0.004406502214260399, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0016344427305739373, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.010447447188198566, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.013430317863821983, + "weighted_rkd_loss": 0.005223723594099283 + }, + { + "epoch": 6.549413735343384, + "flow_matching_loss": 0.0178004146553576, + "grad_norm": 0.16389556229114532, + "learning_rate": 3.7384983613663195e-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": 0.0039194801589474085, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0016993709024973214, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.009538331348448992, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.0178004146553576, + "weighted_rkd_loss": 0.004769165674224496 + }, + { + "epoch": 6.566164154103853, + "flow_matching_loss": 0.014947613794356584, + "grad_norm": 0.202504500746727, + "learning_rate": 3.6732114540536503e-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": 0.004656322370283306, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.001776739489287138, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.011089384276419877, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.014947613794356584, + "weighted_rkd_loss": 0.005544692138209939 + }, + { + "epoch": 6.582914572864322, + "flow_matching_loss": 0.015553992334753275, + "grad_norm": 0.19275301694869995, + "learning_rate": 3.608420017165683e-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": 0.004351094085723162, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0019604666507802905, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.010662654880434275, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.015553992334753275, + "weighted_rkd_loss": 0.005331327440217137 + }, + { + "epoch": 6.599664991624791, + "flow_matching_loss": 0.013738202583044767, + "grad_norm": 0.16886192560195923, + "learning_rate": 3.5441268848897664e-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": 0.004141055117361248, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0016448502079583704, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.009926960431039333, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.013738202583044767, + "weighted_rkd_loss": 0.0049634802155196665 + }, + { + "epoch": 6.61641541038526, + "flow_matching_loss": 0.013459889078512787, + "grad_norm": 0.11751831322908401, + "learning_rate": 3.480334869615792e-06, + "loss": 0.0179, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.0036029696464538576, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0016099985863547773, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.008815937908366323, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.013459889078512787, + "weighted_rkd_loss": 0.004407968954183161 + }, + { + "epoch": 6.633165829145729, + "flow_matching_loss": 0.012848709896206855, + "grad_norm": 0.18301798403263092, + "learning_rate": 3.417046761813155e-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": 0.003821439016610384, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.001900578138884157, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.009543456276878715, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.012848709896206855, + "weighted_rkd_loss": 0.0047717281384393575 + }, + { + "epoch": 6.649916247906198, + "flow_matching_loss": 0.01411162167787552, + "grad_norm": 0.1493939906358719, + "learning_rate": 3.3542653299087263e-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": 0.003471779148094356, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0017120651376899333, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.008655623439699412, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.01411162167787552, + "weighted_rkd_loss": 0.004327811719849706 + }, + { + "epoch": 6.666666666666667, + "flow_matching_loss": 0.010548923909664155, + "grad_norm": 0.14926926791667938, + "learning_rate": 3.2919933201657117e-06, + "loss": 0.0156, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.004148266627453268, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.001756682724226266, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.010053216014057398, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.010548923909664155, + "weighted_rkd_loss": 0.005026608007028699 + }, + { + "epoch": 6.683417085427136, + "flow_matching_loss": 0.014358358830213547, + "grad_norm": 0.1853066384792328, + "learning_rate": 3.2302334565635584e-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": 0.003923210455104709, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.001574484136654064, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.009420905215665697, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.014358358830213547, + "weighted_rkd_loss": 0.004710452607832849 + }, + { + "epoch": 6.700167504187605, + "flow_matching_loss": 0.01270328126847744, + "grad_norm": 0.19508641958236694, + "learning_rate": 3.1689884406787674e-06, + "loss": 0.0172, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.003606516565196216, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0017539363470859826, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.008966969326138496, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.01270328126847744, + "weighted_rkd_loss": 0.004483484663069248 + }, + { + "epoch": 6.716917922948074, + "flow_matching_loss": 0.01392309032380581, + "grad_norm": 0.16966725885868073, + "learning_rate": 3.1082609515667424e-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": 0.0044722165912389755, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0017818786611314863, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.010726311895996331, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.01392309032380581, + "weighted_rkd_loss": 0.005363155947998166 + }, + { + "epoch": 6.733668341708543, + "flow_matching_loss": 0.012681249156594276, + "grad_norm": 0.20630866289138794, + "learning_rate": 3.048053645644594e-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": 0.004192276950925588, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0016123710433021189, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.009996925015002489, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.012681249156594276, + "weighted_rkd_loss": 0.0049984625075012445 + }, + { + "epoch": 6.750418760469012, + "flow_matching_loss": 0.013639771938323974, + "grad_norm": 0.16677002608776093, + "learning_rate": 2.988369156574916e-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": 0.0040628035785630345, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0015570100396871567, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.00968261705711484, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.013639771938323974, + "weighted_rkd_loss": 0.00484130852855742 + }, + { + "epoch": 6.767169179229481, + "flow_matching_loss": 0.01363765113055706, + "grad_norm": 0.12832656502723694, + "learning_rate": 2.9292100951506225e-06, + "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": 0.004342612088657915, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0018042451236397028, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.010489469300955534, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.01363765113055706, + "weighted_rkd_loss": 0.005244734650477767 + }, + { + "epoch": 6.78391959798995, + "flow_matching_loss": 0.013352138735353947, + "grad_norm": 0.15868757665157318, + "learning_rate": 2.8705790491807033e-06, + "loss": 0.0179, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.003720019944012165, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0015638376004062592, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.009003877593204378, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.013352138735353947, + "weighted_rkd_loss": 0.004501938796602189 + }, + { + "epoch": 6.800670016750419, + "flow_matching_loss": 0.012715986743569375, + "grad_norm": 0.18388469517230988, + "learning_rate": 2.8124785833770433e-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": 0.004414536338299513, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0018238211283460259, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01065289373509586, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.012715986743569375, + "weighted_rkd_loss": 0.00532644686754793 + }, + { + "epoch": 6.817420435510888, + "flow_matching_loss": 0.014656269457191229, + "grad_norm": 0.20179112255573273, + "learning_rate": 2.754911239242241e-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": 0.004632821911945939, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0020091856247745456, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.011274829413741827, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.014656269457191229, + "weighted_rkd_loss": 0.005637414706870914 + }, + { + "epoch": 6.834170854271357, + "flow_matching_loss": 0.012329512694850563, + "grad_norm": 0.19481857120990753, + "learning_rate": 2.6978795349584204e-06, + "loss": 0.0179, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.004721343913115561, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.001782020868267864, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01122470865957439, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.012329512694850563, + "weighted_rkd_loss": 0.005612354329787195 + }, + { + "epoch": 6.850921273031826, + "flow_matching_loss": 0.013932113721966743, + "grad_norm": 0.17180730402469635, + "learning_rate": 2.641385965277088e-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": 0.004377597151324153, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0019918588106520472, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.010747053101658821, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.013932113721966743, + "weighted_rkd_loss": 0.005373526550829411 + }, + { + "epoch": 6.867671691792295, + "flow_matching_loss": 0.012642812356352806, + "grad_norm": 0.17917793989181519, + "learning_rate": 2.5854330014099923e-06, + "loss": 0.017, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.0035743531887419524, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0015058233228046447, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.008654529694467783, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.012642812356352806, + "weighted_rkd_loss": 0.0043272648472338915 + }, + { + "epoch": 6.884422110552764, + "flow_matching_loss": 0.01389442621730268, + "grad_norm": 0.12711286544799805, + "learning_rate": 2.5300230909210298e-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": 0.0035649200435727835, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0015694846806582063, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.008699324633926152, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.01389442621730268, + "weighted_rkd_loss": 0.004349662316963076 + }, + { + "epoch": 6.901172529313233, + "flow_matching_loss": 0.012807622458785773, + "grad_norm": 0.19579429924488068, + "learning_rate": 2.4751586576191983e-06, + "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": 0.004366207169368863, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0018814566778019071, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01061387094669044, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.012807622458785773, + "weighted_rkd_loss": 0.00530693547334522 + }, + { + "epoch": 6.917922948073702, + "flow_matching_loss": 0.012661585444584489, + "grad_norm": 0.1477232426404953, + "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": 0.004671722766943276, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0016743403277359903, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01101778601296246, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.012661585444584489, + "weighted_rkd_loss": 0.00550889300648123 + }, + { + "epoch": 6.934673366834171, + "flow_matching_loss": 0.012656646966934203, + "grad_norm": 0.17730483412742615, + "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": 0.004852761048823595, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0018881546217016875, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.011593676824122667, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.012656646966934203, + "weighted_rkd_loss": 0.005796838412061333 + }, + { + "epoch": 6.95142378559464, + "flow_matching_loss": 0.011845463747158646, + "grad_norm": 0.1716385930776596, + "learning_rate": 2.313862100383404e-06, + "loss": 0.0164, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.0035892325220629575, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0018717827624641359, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.009050247818231582, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.011845463747158646, + "weighted_rkd_loss": 0.004525123909115791 + }, + { + "epoch": 6.968174204355109, + "flow_matching_loss": 0.012647939100861549, + "grad_norm": 0.1931212693452835, + "learning_rate": 2.2612033351327406e-06, + "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": 0.004044805024750531, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0015366453211754561, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.009626255370676517, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.012647939100861549, + "weighted_rkd_loss": 0.004813127685338259 + }, + { + "epoch": 6.984924623115578, + "flow_matching_loss": 0.01442919559776783, + "grad_norm": 0.19145072996616364, + "learning_rate": 2.2091018061161788e-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": 0.0043393786996603016, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.002081942302174866, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.010760699631646276, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.01442919559776783, + "weighted_rkd_loss": 0.005380349815823138 + }, + { + "epoch": 7.001675041876047, + "flow_matching_loss": 0.013325535040348769, + "grad_norm": 0.16990314424037933, + "learning_rate": 2.15755979242338e-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": 0.00459274286404252, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.00216435887850821, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.011349844746291637, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.013325535040348769, + "weighted_rkd_loss": 0.005674922373145819 + }, + { + "epoch": 7.018425460636516, + "flow_matching_loss": 0.015028421999886632, + "grad_norm": 0.1645367592573166, + "learning_rate": 2.106579548669e-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": 0.00421563507989049, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.001719225279521197, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.010150495450943708, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.015028421999886632, + "weighted_rkd_loss": 0.005075247725471854 + }, + { + "epoch": 7.035175879396985, + "flow_matching_loss": 0.011063610622659325, + "grad_norm": 0.14495235681533813, + "learning_rate": 2.0561633048940477e-06, + "loss": 0.0154, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.0035904986783862115, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0014928206335753202, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.008673817990347743, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.011063610622659325, + "weighted_rkd_loss": 0.004336908995173872 + }, + { + "epoch": 7.051926298157454, + "flow_matching_loss": 0.014337676949799061, + "grad_norm": 0.14312154054641724, + "learning_rate": 2.006313266468355e-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": 0.0042888281866908075, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0018349151243455708, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.010412571532651783, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.014337676949799061, + "weighted_rkd_loss": 0.005206285766325891 + }, + { + "epoch": 7.068676716917923, + "flow_matching_loss": 0.011880451068282127, + "grad_norm": 0.17340722680091858, + "learning_rate": 1.957031613994097e-06, + "loss": 0.0163, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.0037391659570857882, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0014357647683937102, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.008914096700027584, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.011880451068282127, + "weighted_rkd_loss": 0.004457048350013792 + }, + { + "epoch": 7.085427135678392, + "flow_matching_loss": 0.013384684594348073, + "grad_norm": 0.14712361991405487, + "learning_rate": 1.9083205032103995e-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": 0.0043742188951000575, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0020250638597644867, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01077350163832307, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.013384684594348073, + "weighted_rkd_loss": 0.005386750819161535 + }, + { + "epoch": 7.102177554438861, + "flow_matching_loss": 0.014141054265201092, + "grad_norm": 0.14053156971931458, + "learning_rate": 1.8601820648990536e-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": 0.003534198086708784, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0016627753793727607, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.008731171535328031, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.014141054265201092, + "weighted_rkd_loss": 0.004365585767664016 + }, + { + "epoch": 7.11892797319933, + "flow_matching_loss": 0.012220955546945333, + "grad_norm": 0.14366623759269714, + "learning_rate": 1.8126184047912936e-06, + "loss": 0.0174, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.0042095793643966315, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.002016812388319522, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.010435971105471254, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.012220955546945333, + "weighted_rkd_loss": 0.005217985552735627 + }, + { + "epoch": 7.135678391959799, + "flow_matching_loss": 0.012571109691634774, + "grad_norm": 0.14746060967445374, + "learning_rate": 1.7656316034757024e-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": 0.004236602829769254, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0017786037409678102, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.010251809330657124, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.012571109691634774, + "weighted_rkd_loss": 0.005125904665328562 + }, + { + "epoch": 7.152428810720268, + "flow_matching_loss": 0.013217747118324042, + "grad_norm": 0.16644246876239777, + "learning_rate": 1.7192237163071816e-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": 0.004921759618446231, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0020453046192415057, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.011888823937624693, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.013217747118324042, + "weighted_rkd_loss": 0.0059444119688123465 + }, + { + "epoch": 7.169179229480737, + "flow_matching_loss": 0.013593366229906678, + "grad_norm": 0.1624123603105545, + "learning_rate": 1.6733967733170573e-06, + "loss": 0.0179, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.00354135453235358, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0015589617833029479, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.008641670923680067, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.013593366229906678, + "weighted_rkd_loss": 0.0043208354618400335 + }, + { + "epoch": 7.185929648241206, + "flow_matching_loss": 0.011336163431406022, + "grad_norm": 0.15716801583766937, + "learning_rate": 1.628152779124279e-06, + "loss": 0.0166, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.004345083329826594, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.001911070611095056, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.010601237136870622, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.011336163431406022, + "weighted_rkd_loss": 0.005300618568435311 + }, + { + "epoch": 7.202680067001675, + "flow_matching_loss": 0.012439127638936042, + "grad_norm": 0.1175704151391983, + "learning_rate": 1.583493712847715e-06, + "loss": 0.0172, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.003965030517429113, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0015500714420340956, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.009480132395401597, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.012439127638936042, + "weighted_rkd_loss": 0.004740066197700799 + }, + { + "epoch": 7.219430485762144, + "flow_matching_loss": 0.01372097311541438, + "grad_norm": 0.14474470913410187, + "learning_rate": 1.5394215280196084e-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": 0.004111591842956841, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0015248423907905817, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.009748026169836521, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.01372097311541438, + "weighted_rkd_loss": 0.004874013084918261 + }, + { + "epoch": 7.236180904522613, + "flow_matching_loss": 0.012772003002464772, + "grad_norm": 0.15785634517669678, + "learning_rate": 1.4959381525000904e-06, + "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": 0.004256247635930776, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.001629356446210295, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.010141851706430316, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.012772003002464772, + "weighted_rkd_loss": 0.005070925853215158 + }, + { + "epoch": 7.252931323283082, + "flow_matching_loss": 0.014338200632482766, + "grad_norm": 0.160471573472023, + "learning_rate": 1.453045488392879e-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": 0.004117710143327713, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0016662140609696508, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.009901634324342013, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.014338200632482766, + "weighted_rkd_loss": 0.004950817162171006 + }, + { + "epoch": 7.269681742043551, + "flow_matching_loss": 0.012580685364082456, + "grad_norm": 0.13657648861408234, + "learning_rate": 1.4107454119620484e-06, + "loss": 0.0169, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.0036080827238038184, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0013286747212987392, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.008544840198010207, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.012580685364082456, + "weighted_rkd_loss": 0.0042724200990051035 + }, + { + "epoch": 7.28643216080402, + "flow_matching_loss": 0.013763728784397244, + "grad_norm": 0.16480371356010437, + "learning_rate": 1.3690397735499778e-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": 0.004622260120231658, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0022792898351326587, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.011523810215294361, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.013763728784397244, + "weighted_rkd_loss": 0.0057619051076471806 + }, + { + "epoch": 7.303182579564489, + "flow_matching_loss": 0.014634063653647899, + "grad_norm": 0.1543356031179428, + "learning_rate": 1.327930397496403e-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": 0.004943400342017412, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0021274060709401965, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01201420659199357, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.014634063653647899, + "weighted_rkd_loss": 0.006007103295996785 + }, + { + "epoch": 7.319932998324958, + "flow_matching_loss": 0.011182602029293776, + "grad_norm": 0.23041939735412598, + "learning_rate": 1.287419082058604e-06, + "loss": 0.0161, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.003948941268026829, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0018433966557495295, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.009741279110312463, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.011182602029293776, + "weighted_rkd_loss": 0.004870639555156231 + }, + { + "epoch": 7.336683417085427, + "flow_matching_loss": 0.012292292807251215, + "grad_norm": 0.1963876187801361, + "learning_rate": 1.2475075993327589e-06, + "loss": 0.0169, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.003892730991356075, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.001485452667111531, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.009270914644002915, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.012292292807251215, + "weighted_rkd_loss": 0.0046354573220014576 + }, + { + "epoch": 7.353433835845896, + "flow_matching_loss": 0.014200325449928641, + "grad_norm": 0.19687843322753906, + "learning_rate": 1.208197695176413e-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": 0.00432422065641731, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.002694979298394173, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01134342052973807, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.014200325449928641, + "weighted_rkd_loss": 0.005671710264869035 + }, + { + "epoch": 7.370184254606365, + "flow_matching_loss": 0.011306853778660298, + "grad_norm": 0.16218173503875732, + "learning_rate": 1.1694910891321259e-06, + "loss": 0.0164, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.004136376339010895, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.001833427872043103, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.010106180747970938, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.011306853778660298, + "weighted_rkd_loss": 0.005053090373985469 + }, + { + "epoch": 7.386934673366834, + "flow_matching_loss": 0.013065098226070404, + "grad_norm": 0.1709485948085785, + "learning_rate": 1.1313894743522363e-06, + "loss": 0.0174, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.0036178565234877167, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0015012272633612156, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.008736940333619714, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.013065098226070404, + "weighted_rkd_loss": 0.004368470166809857 + }, + { + "epoch": 7.403685092127303, + "flow_matching_loss": 0.01347271353006363, + "grad_norm": 0.1800774782896042, + "learning_rate": 1.0938945175248016e-06, + "loss": 0.0179, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.003690591617487371, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0015590842347592115, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.00894026760943234, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.01347271353006363, + "weighted_rkd_loss": 0.00447013380471617 + }, + { + "epoch": 7.420435510887772, + "flow_matching_loss": 0.013602646300569177, + "grad_norm": 0.1115361675620079, + "learning_rate": 1.0570078588007086e-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": 0.0041832276852801446, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0017717199341859668, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.010138175310567021, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.013602646300569177, + "weighted_rkd_loss": 0.0050690876552835105 + }, + { + "epoch": 7.437185929648241, + "flow_matching_loss": 0.011779169458895922, + "grad_norm": 0.13396503031253815, + "learning_rate": 1.0207311117219009e-06, + "loss": 0.0168, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.004144427995197475, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.001768887892831117, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.010057743871584535, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.011779169458895922, + "weighted_rkd_loss": 0.005028871935792267 + }, + { + "epoch": 7.45393634840871, + "flow_matching_loss": 0.011645021988078952, + "grad_norm": 0.11680609732866287, + "learning_rate": 9.850658631508198e-07, + "loss": 0.0161, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.003669659234583378, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0015738890622742474, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.008913207612931729, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.011645021988078952, + "weighted_rkd_loss": 0.0044566038064658645 + }, + { + "epoch": 7.4706867671691795, + "flow_matching_loss": 0.012487611640244723, + "grad_norm": 0.15301018953323364, + "learning_rate": 9.500136732009751e-07, + "loss": 0.0169, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.0036924283020198347, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.001380647497717291, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.008765504136681557, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.012487611640244723, + "weighted_rkd_loss": 0.0043827520683407785 + }, + { + "epoch": 7.4874371859296485, + "flow_matching_loss": 0.01187393870204687, + "grad_norm": 0.13816434144973755, + "learning_rate": 9.155760751687054e-07, + "loss": 0.0162, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.0035726704518310726, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0014808060775976628, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.008626147080212832, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.01187393870204687, + "weighted_rkd_loss": 0.004313073540106416 + }, + { + "epoch": 7.5041876046901175, + "flow_matching_loss": 0.011787522397935391, + "grad_norm": 0.1930665522813797, + "learning_rate": 8.817545754661155e-07, + "loss": 0.0165, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.0038458931958302855, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0016517811745870858, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.009343567583709956, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.011787522397935391, + "weighted_rkd_loss": 0.004671783791854978 + }, + { + "epoch": 7.5209380234505865, + "flow_matching_loss": 0.014746030792593956, + "grad_norm": 0.15736708045005798, + "learning_rate": 8.485506535551668e-07, + "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": 0.004147592210210859, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0017917696095537394, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.010086953965947032, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.014746030792593956, + "weighted_rkd_loss": 0.005043476982973516 + }, + { + "epoch": 7.5376884422110555, + "flow_matching_loss": 0.011041402723640204, + "grad_norm": 0.15687477588653564, + "learning_rate": 8.159657618829685e-07, + "loss": 0.0155, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.003685565129853785, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0015495843777898698, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.008920714585110545, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.011041402723640204, + "weighted_rkd_loss": 0.004460357292555272 + }, + { + "epoch": 7.5544388609715245, + "flow_matching_loss": 0.013714790437370538, + "grad_norm": 0.12664705514907837, + "learning_rate": 7.840013258182471e-07, + "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": 0.0038969279499724506, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0014690222684293984, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.0092628781683743, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.013714790437370538, + "weighted_rkd_loss": 0.00463143908418715 + }, + { + "epoch": 7.5711892797319935, + "flow_matching_loss": 0.014138914924114943, + "grad_norm": 0.12797820568084717, + "learning_rate": 7.526587435889809e-07, + "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": 0.0034030490089207886, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0013546278292778879, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.008160725887864827, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.014138914924114943, + "weighted_rkd_loss": 0.004080362943932414 + }, + { + "epoch": 7.5879396984924625, + "flow_matching_loss": 0.012961727753281593, + "grad_norm": 0.15495085716247559, + "learning_rate": 7.219393862212593e-07, + "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": 0.003796363272704184, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.001647556375246495, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.009240282978862525, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.012961727753281593, + "weighted_rkd_loss": 0.004620141489431262 + }, + { + "epoch": 7.6046901172529315, + "flow_matching_loss": 0.01382149439305067, + "grad_norm": 0.17086490988731384, + "learning_rate": 6.918445974792819e-07, + "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": 0.003138877125456929, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0015061549376696349, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.007783909235149622, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.01382149439305067, + "weighted_rkd_loss": 0.003891954617574811 + }, + { + "epoch": 7.6214405360134005, + "flow_matching_loss": 0.01289975387044251, + "grad_norm": 0.10146980732679367, + "learning_rate": 6.6237569380661e-07, + "loss": 0.017, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.0034003766952082513, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0013089508167468012, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.008109704172238708, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.01289975387044251, + "weighted_rkd_loss": 0.004054852086119354 + }, + { + "epoch": 7.63819095477387, + "flow_matching_loss": 0.013388771284371614, + "grad_norm": 0.19025149941444397, + "learning_rate": 6.335339642685495e-07, + "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": 0.0039990865625441074, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.00168166448129341, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.009679837571457028, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.013388771284371614, + "weighted_rkd_loss": 0.004839918785728514 + }, + { + "epoch": 7.654941373534339, + "flow_matching_loss": 0.013220328884199262, + "grad_norm": 0.1995600312948227, + "learning_rate": 6.05320670495787e-07, + "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": 0.0038794698659330605, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0017690667242277413, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.009528006473556162, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.013220328884199262, + "weighted_rkd_loss": 0.004764003236778081 + }, + { + "epoch": 7.671691792294808, + "flow_matching_loss": 0.012908745370805264, + "grad_norm": 0.18338993191719055, + "learning_rate": 5.777370466291954e-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": 0.004892875347286463, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.002224490814842284, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.012010241625830531, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.012908745370805264, + "weighted_rkd_loss": 0.0060051208129152656 + }, + { + "epoch": 7.688442211055277, + "flow_matching_loss": 0.012604714278131724, + "grad_norm": 0.17824886739253998, + "learning_rate": 5.507842992658336e-07, + "loss": 0.017, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.003669789177365601, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0014661247783806174, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.008805703138932586, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.012604714278131724, + "weighted_rkd_loss": 0.004402851569466293 + }, + { + "epoch": 7.705192629815746, + "flow_matching_loss": 0.013908560667186976, + "grad_norm": 0.1996762901544571, + "learning_rate": 5.24463607406192e-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": 0.004282183572649955, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0017393660498782992, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.010303733311593533, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.013908560667186976, + "weighted_rkd_loss": 0.0051518666557967665 + }, + { + "epoch": 7.721943048576215, + "flow_matching_loss": 0.011516436655074357, + "grad_norm": 0.15033936500549316, + "learning_rate": 4.987761224025944e-07, + "loss": 0.0165, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.004200274380855263, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0015304363565519451, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.00993098528124392, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.011516436655074357, + "weighted_rkd_loss": 0.00496549264062196 + }, + { + "epoch": 7.738693467336684, + "flow_matching_loss": 0.016077415226027368, + "grad_norm": 0.15527673065662384, + "learning_rate": 4.737229679088556e-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": 0.0033825806342065333, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0015490434423554688, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.008314204728230834, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.016077415226027368, + "weighted_rkd_loss": 0.004157102364115417 + }, + { + "epoch": 7.755443886097153, + "flow_matching_loss": 0.013937278091907502, + "grad_norm": 0.15270797908306122, + "learning_rate": 4.493052398311115e-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": 0.0038180277915671468, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0016788874345365912, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.009314942918717862, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.013937278091907502, + "weighted_rkd_loss": 0.004657471459358931 + }, + { + "epoch": 7.772194304857622, + "flow_matching_loss": 0.012719677481800317, + "grad_norm": 0.160204216837883, + "learning_rate": 4.2552400627989043e-07, + "loss": 0.0173, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.0037694335216656326, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0015906929154880344, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.00912955985404551, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.012719677481800317, + "weighted_rkd_loss": 0.004564779927022755 + }, + { + "epoch": 7.788944723618091, + "flow_matching_loss": 0.012721182778477669, + "grad_norm": 0.18550850450992584, + "learning_rate": 4.023803075233923e-07, + "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": 0.004492715140804649, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0019040567451156676, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01088948706164956, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.012721182778477669, + "weighted_rkd_loss": 0.00544474353082478 + }, + { + "epoch": 7.80569514237856, + "flow_matching_loss": 0.013466277066618204, + "grad_norm": 0.14704406261444092, + "learning_rate": 3.7987515594196997e-07, + "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": 0.0033923544222489, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0013608533889055252, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.008145562140271068, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.013466277066618204, + "weighted_rkd_loss": 0.004072781070135534 + }, + { + "epoch": 7.822445561139029, + "flow_matching_loss": 0.013312026532366872, + "grad_norm": 0.21852347254753113, + "learning_rate": 3.580095359838648e-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": 0.004508821619674563, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0016108532319776714, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.010628496576100589, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.013312026532366872, + "weighted_rkd_loss": 0.005314248288050294 + }, + { + "epoch": 7.839195979899498, + "flow_matching_loss": 0.012858473043888808, + "grad_norm": 0.17716294527053833, + "learning_rate": 3.3678440412212576e-07, + "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": 0.004035146767273545, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.001507147657684982, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.009577441122382879, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.012858473043888808, + "weighted_rkd_loss": 0.0047887205611914395 + }, + { + "epoch": 7.855946398659967, + "flow_matching_loss": 0.014230535924434662, + "grad_norm": 0.13805650174617767, + "learning_rate": 3.162006888127833e-07, + "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": 0.004414458805695176, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0018795684911310673, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.010708486149087548, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.014230535924434662, + "weighted_rkd_loss": 0.005354243074543774 + }, + { + "epoch": 7.872696817420436, + "flow_matching_loss": 0.014107253961265087, + "grad_norm": 0.16493017971515656, + "learning_rate": 2.9625929045422327e-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": 0.0039233009098097685, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0015600109356455505, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.00940661272034049, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.014107253961265087, + "weighted_rkd_loss": 0.004703306360170245 + }, + { + "epoch": 7.889447236180905, + "flow_matching_loss": 0.015400014398619533, + "grad_norm": 0.16884583234786987, + "learning_rate": 2.7696108134781186e-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": 0.0032546573784202337, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0012738020333927125, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.00778311681933701, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.015400014398619533, + "weighted_rkd_loss": 0.003891558409668505 + }, + { + "epoch": 7.906197654941374, + "flow_matching_loss": 0.013906987011432647, + "grad_norm": 0.1894207000732422, + "learning_rate": 2.583069056597376e-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": 0.004085881053470075, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0017150958010461181, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.009886857867240906, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.013906987011432647, + "weighted_rkd_loss": 0.004943428933620453 + }, + { + "epoch": 7.922948073701843, + "flow_matching_loss": 0.011825017677620054, + "grad_norm": 0.16371731460094452, + "learning_rate": 2.4029757938407116e-07, + "loss": 0.0171, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.004440854152198881, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0017637885699514299, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01064549693837762, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.011825017677620054, + "weighted_rkd_loss": 0.00532274846918881 + }, + { + "epoch": 7.939698492462312, + "flow_matching_loss": 0.014789005741477013, + "grad_norm": 0.13796105980873108, + "learning_rate": 2.229338903070871e-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": 0.003778302017599344, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0013750840036664159, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.008931688079610468, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.014789005741477013, + "weighted_rkd_loss": 0.004465844039805234 + }, + { + "epoch": 7.956448911222781, + "flow_matching_loss": 0.013503665756434202, + "grad_norm": 0.13203434646129608, + "learning_rate": 2.0621659797279645e-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": 0.004043603711761534, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0016320106165949254, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.009719217987731099, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.013503665756434202, + "weighted_rkd_loss": 0.004859608993865549 + }, + { + "epoch": 7.97319932998325, + "flow_matching_loss": 0.01366380015388131, + "grad_norm": 0.18163886666297913, + "learning_rate": 1.901464336497233e-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": 0.0043127588927745816, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.001717785291839391, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.010343303112313152, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.01366380015388131, + "weighted_rkd_loss": 0.005171651556156576 + }, + { + "epoch": 7.989949748743719, + "flow_matching_loss": 0.012864657817408443, + "grad_norm": 0.17755556106567383, + "learning_rate": 1.7472410029891527e-07, + "loss": 0.0174, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.0037821149686351418, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0014765637111850082, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.009040793776512146, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.012864657817408443, + "weighted_rkd_loss": 0.004520396888256073 + }, + { + "epoch": 8.006700167504187, + "flow_matching_loss": 0.011925599724054336, + "grad_norm": 0.20086751878261566, + "learning_rate": 1.5995027254319173e-07, + "loss": 0.0161, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.0034393739653751252, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0014759833575226367, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.008354731323197483, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.011925599724054336, + "weighted_rkd_loss": 0.004177365661598742 + }, + { + "epoch": 8.023450586264657, + "flow_matching_loss": 0.013343129586428405, + "grad_norm": 0.12880833446979523, + "learning_rate": 1.4582559663764372e-07, + "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": 0.0033495205221697687, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0017303403757978231, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.00842938139103353, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.013343129586428405, + "weighted_rkd_loss": 0.004214690695516765 + }, + { + "epoch": 8.040201005025125, + "flow_matching_loss": 0.015160613507032395, + "grad_norm": 0.16944606602191925, + "learning_rate": 1.323506904413485e-07, + "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": 0.0036136694019660355, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.001553205365780741, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.00878054415807128, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.015160613507032395, + "weighted_rkd_loss": 0.00439027207903564 + }, + { + "epoch": 8.056951423785595, + "flow_matching_loss": 0.012301782937720418, + "grad_norm": 0.21078728139400482, + "learning_rate": 1.195261433903594e-07, + "loss": 0.0172, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.004058028594590724, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0015868473798036574, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.009702904568985104, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.012301782937720418, + "weighted_rkd_loss": 0.004851452284492552 + }, + { + "epoch": 8.073701842546063, + "flow_matching_loss": 0.012389839068055153, + "grad_norm": 0.141947403550148, + "learning_rate": 1.0735251647190969e-07, + "loss": 0.017, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.0038053673692047597, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.001513481535948813, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.009124216344207525, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.012389839068055153, + "weighted_rkd_loss": 0.004562108172103762 + }, + { + "epoch": 8.090452261306533, + "flow_matching_loss": 0.012021586624905467, + "grad_norm": 0.17158496379852295, + "learning_rate": 9.583034219987408e-08, + "loss": 0.0164, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.003686944651417434, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0014448780042584986, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.008818767266348004, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.012021586624905467, + "weighted_rkd_loss": 0.004409383633174002 + }, + { + "epoch": 8.107202680067001, + "flow_matching_loss": 0.010950662288814783, + "grad_norm": 0.12545715272426605, + "learning_rate": 8.496012459148717e-08, + "loss": 0.016, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.004086814215406775, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0019535994331818076, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01012722784653306, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.010950662288814783, + "weighted_rkd_loss": 0.00506361392326653 + }, + { + "epoch": 8.12395309882747, + "flow_matching_loss": 0.015066776238381863, + "grad_norm": 0.14003939926624298, + "learning_rate": 7.474233914527618e-08, + "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": 0.0033574032597243784, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0013386573467869312, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.008053463837131857, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.015066776238381863, + "weighted_rkd_loss": 0.004026731918565929 + }, + { + "epoch": 8.140703517587939, + "flow_matching_loss": 0.01541524352505803, + "grad_norm": 0.10983029007911682, + "learning_rate": 6.517743282028254e-08, + "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": 0.003956313151866198, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0016391688433941454, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.009551795152947306, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.01541524352505803, + "weighted_rkd_loss": 0.004775897576473653 + }, + { + "epoch": 8.157453936348409, + "flow_matching_loss": 0.0127577087841928, + "grad_norm": 0.145966574549675, + "learning_rate": 5.6265824016492604e-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": 0.004257415048778057, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0016393918893299996, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01015422185882926, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.0127577087841928, + "weighted_rkd_loss": 0.00507711092941463 + }, + { + "epoch": 8.174204355108877, + "flow_matching_loss": 0.0156515721231699, + "grad_norm": 0.12217182666063309, + "learning_rate": 4.800790255654719e-08, + "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": 0.003486564732156694, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0016595104534644634, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.008632639842107891, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.0156515721231699, + "weighted_rkd_loss": 0.004316319921053946 + }, + { + "epoch": 8.190954773869347, + "flow_matching_loss": 0.013703823648393154, + "grad_norm": 0.13512387871742249, + "learning_rate": 4.040402966867862e-08, + "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": 0.004013277473859489, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0014476361509878188, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.009474191255867482, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.013703823648393154, + "weighted_rkd_loss": 0.004737095627933741 + }, + { + "epoch": 8.207705192629815, + "flow_matching_loss": 0.011992666684091092, + "grad_norm": 0.15322472155094147, + "learning_rate": 3.3454537970925014e-08, + "loss": 0.0169, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.003927948325872422, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0020091683545615524, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.009865065244957805, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.011992666684091092, + "weighted_rkd_loss": 0.0049325326224789025 + }, + { + "epoch": 8.224455611390285, + "flow_matching_loss": 0.015503291366621852, + "grad_norm": 0.1615697145462036, + "learning_rate": 2.715973145656192e-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": 0.0033490344532765446, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.001463252492249012, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.008161321375519037, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.015503291366621852, + "weighted_rkd_loss": 0.0040806606877595184 + }, + { + "epoch": 8.241206030150753, + "flow_matching_loss": 0.012654225155711175, + "grad_norm": 0.13093577325344086, + "learning_rate": 2.151988548082129e-08, + "loss": 0.0168, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.00344906086102128, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0013728271587751806, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.008270948939025403, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.012654225155711175, + "weighted_rkd_loss": 0.004135474469512701 + }, + { + "epoch": 8.257956448911223, + "flow_matching_loss": 0.013267476949840784, + "grad_norm": 0.18575097620487213, + "learning_rate": 1.6535246748837772e-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": 0.004015879752114415, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.002008393610594794, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.010040153004229068, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.013267476949840784, + "weighted_rkd_loss": 0.005020076502114534 + }, + { + "epoch": 8.274706867671691, + "flow_matching_loss": 0.013465237710624934, + "grad_norm": 0.1440298706293106, + "learning_rate": 1.2206033304855702e-08, + "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": 0.003453168785199523, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0014610151178203524, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.008367352746427059, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.013465237710624934, + "weighted_rkd_loss": 0.004183676373213529 + }, + { + "epoch": 8.291457286432161, + "flow_matching_loss": 0.012969103502109646, + "grad_norm": 0.12818564474582672, + "learning_rate": 8.532434522695031e-09, + "loss": 0.0179, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.004048713785596192, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0016746725188568235, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.009772100113332271, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.012969103502109646, + "weighted_rkd_loss": 0.004886050056666136 + }, + { + "epoch": 8.308207705192629, + "flow_matching_loss": 0.013686891552060843, + "grad_norm": 0.15679726004600525, + "learning_rate": 5.5146110974679805e-09, + "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": 0.0036764851538464427, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0014912928687408566, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.008844263153150678, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.013686891552060843, + "weighted_rkd_loss": 0.004422131576575339 + }, + { + "epoch": 8.324958123953099, + "flow_matching_loss": 0.01356009249575436, + "grad_norm": 0.1348220556974411, + "learning_rate": 3.1526950385496467e-09, + "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": 0.003303223033435643, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0013120193034410477, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.007918465510010719, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.01356009249575436, + "weighted_rkd_loss": 0.0039592327550053595 + }, + { + "epoch": 8.341708542713567, + "flow_matching_loss": 0.010772908199578523, + "grad_norm": 0.14952634274959564, + "learning_rate": 1.4467896637959666e-09, + "loss": 0.0152, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.003719139681197703, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0014399453764781355, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.008878224762156606, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.010772908199578523, + "weighted_rkd_loss": 0.004439112381078303 + }, + { + "epoch": 8.358458961474037, + "flow_matching_loss": 0.012926962599158287, + "grad_norm": 0.12651483714580536, + "learning_rate": 3.9696959503898733e-10, + "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": 0.003811643342487514, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0014979283267166465, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.009121215017512441, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.012926962599158287, + "weighted_rkd_loss": 0.004560607508756221 + }, + { + "epoch": 8.375209380234505, + "flow_matching_loss": 0.011858177138492465, + "grad_norm": 0.20581090450286865, + "learning_rate": 3.280754806156416e-12, + "loss": 0.017, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.004163221316412091, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0019380579411517828, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.010264500416815281, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.011858177138492465, + "weighted_rkd_loss": 0.005132250208407641 + } + ], + "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/seed1/default/train_5shot_seed1_fullactionhead_5k/config.json b/FewShot/39tasks/5shot/FromRKD-Timewarp_FullactionHead/seed1/default/train_5shot_seed1_fullactionhead_5k/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/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/model-00001-of-00002.safetensors b/FewShot/39tasks/5shot/FromRKD-Timewarp_FullactionHead/seed1/default/train_5shot_seed1_fullactionhead_5k/model-00001-of-00002.safetensors new file mode 100644 index 0000000000000000000000000000000000000000..46d8e7b66bb949d86548e5d76fb38ce09b265acf --- /dev/null +++ b/FewShot/39tasks/5shot/FromRKD-Timewarp_FullactionHead/seed1/default/train_5shot_seed1_fullactionhead_5k/model-00001-of-00002.safetensors @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:d3cb77271d1a9fce833f206b6913da1e083de735183e622ded352d7b12681fb1 +size 4999367032 diff --git a/FewShot/39tasks/5shot/FromRKD-Timewarp_FullactionHead/seed1/default/train_5shot_seed1_fullactionhead_5k/model-00002-of-00002.safetensors b/FewShot/39tasks/5shot/FromRKD-Timewarp_FullactionHead/seed1/default/train_5shot_seed1_fullactionhead_5k/model-00002-of-00002.safetensors new file mode 100644 index 0000000000000000000000000000000000000000..3c984996137bf66d92f145056d386104dba4a503 --- /dev/null +++ b/FewShot/39tasks/5shot/FromRKD-Timewarp_FullactionHead/seed1/default/train_5shot_seed1_fullactionhead_5k/model-00002-of-00002.safetensors @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:8e1d911f2e4331b03e74d36773631a8e6ac5b184267257c9c86e233b51cf80f3 +size 2664540348 diff --git a/FewShot/39tasks/5shot/FromRKD-Timewarp_FullactionHead/seed1/default/train_5shot_seed1_fullactionhead_5k/model.safetensors.index.json b/FewShot/39tasks/5shot/FromRKD-Timewarp_FullactionHead/seed1/default/train_5shot_seed1_fullactionhead_5k/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/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/trainer_state.json b/FewShot/39tasks/5shot/FromRKD-Timewarp_FullactionHead/seed1/default/train_5shot_seed1_fullactionhead_5k/trainer_state.json new file mode 100644 index 0000000000000000000000000000000000000000..a170fb3898ee6bb25c4882c6dd4608e6fa4c92ab --- /dev/null +++ b/FewShot/39tasks/5shot/FromRKD-Timewarp_FullactionHead/seed1/default/train_5shot_seed1_fullactionhead_5k/trainer_state.json @@ -0,0 +1,23043 @@ +{ + "best_global_step": null, + "best_metric": null, + "best_model_checkpoint": null, + "epoch": 8.375209380234505, + "eval_steps": 500, + "global_step": 5000, + "is_hyper_param_search": false, + "is_local_process_zero": true, + "is_world_process_zero": true, + "log_history": [ + { + "epoch": 0.01675041876046901, + "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 + }, + { + "epoch": 5.041876046901173, + "flow_matching_loss": 0.015169484168291092, + "grad_norm": 0.21513906121253967, + "learning_rate": 1.1231228239056467e-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": 0.005936745693907142, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0022759507060982286, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01414944203570485, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.015169484168291092, + "weighted_rkd_loss": 0.007074721017852425 + }, + { + "epoch": 5.058626465661642, + "flow_matching_loss": 0.016462551802396773, + "grad_norm": 0.1520560383796692, + "learning_rate": 1.113528559434845e-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": 0.0047940189950168135, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0019913177471607923, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.011579355597496033, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.016462551802396773, + "weighted_rkd_loss": 0.0057896777987480165 + }, + { + "epoch": 5.075376884422111, + "flow_matching_loss": 0.017612619884312154, + "grad_norm": 0.17633816599845886, + "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": 0.005207244050689041, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0022622236749157308, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.012676711846143007, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.017612619884312154, + "weighted_rkd_loss": 0.006338355923071503 + }, + { + "epoch": 5.09212730318258, + "flow_matching_loss": 0.015209727641195058, + "grad_norm": 0.1599225252866745, + "learning_rate": 1.0943911659763612e-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": 0.005452072992920875, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0021627967827953396, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.013066942524164914, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.015209727641195058, + "weighted_rkd_loss": 0.006533471262082457 + }, + { + "epoch": 5.108877721943049, + "flow_matching_loss": 0.014849191531538963, + "grad_norm": 0.15893946588039398, + "learning_rate": 1.084848874120266e-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": 0.005263407132588327, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0025313270860351624, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.013058141246438027, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.014849191531538963, + "weighted_rkd_loss": 0.006529070623219013 + }, + { + "epoch": 5.125628140703517, + "flow_matching_loss": 0.01759036611765623, + "grad_norm": 0.22340935468673706, + "learning_rate": 1.0753247423193196e-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": 0.005642558354884386, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.002568938955664635, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.013854055665433407, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.01759036611765623, + "weighted_rkd_loss": 0.006927027832716703 + }, + { + "epoch": 5.142378559463987, + "flow_matching_loss": 0.015588127821683884, + "grad_norm": 0.1604069173336029, + "learning_rate": 1.0658191871899012e-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": 0.005148043041117489, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.002302269684150815, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.012598355766385794, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.015588127821683884, + "weighted_rkd_loss": 0.006299177883192897 + }, + { + "epoch": 5.159128978224456, + "flow_matching_loss": 0.01548425005748868, + "grad_norm": 0.17967723309993744, + "learning_rate": 1.0563326245357852e-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": 0.004702624608762562, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0019607308669947086, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.011365980049595237, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.01548425005748868, + "weighted_rkd_loss": 0.0056829900247976186 + }, + { + "epoch": 5.175879396984925, + "flow_matching_loss": 0.014517775271087885, + "grad_norm": 0.20446251332759857, + "learning_rate": 1.0468654693299557e-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": 0.00531402169726789, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0026462510228157043, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.013274294324219227, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.014517775271087885, + "weighted_rkd_loss": 0.006637147162109613 + }, + { + "epoch": 5.192629815745394, + "flow_matching_loss": 0.01355633307248354, + "grad_norm": 0.20802079141139984, + "learning_rate": 1.0374181356964501e-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": 0.006404514703899622, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0025749327382072807, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.015383962262421846, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.01355633307248354, + "weighted_rkd_loss": 0.007691981131210923 + }, + { + "epoch": 5.209380234505863, + "flow_matching_loss": 0.014503863453865052, + "grad_norm": 0.18346713483333588, + "learning_rate": 1.0279910368922465e-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": 0.005380265135318041, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0024684223462827504, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.0132289526052773, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.014503863453865052, + "weighted_rkd_loss": 0.00661447630263865 + }, + { + "epoch": 5.226130653266332, + "flow_matching_loss": 0.013800499308854342, + "grad_norm": 0.1441630870103836, + "learning_rate": 1.018584585289186e-05, + "loss": 0.0197, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.004836055263876915, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0021854664431884886, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.011857577040791512, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.013800499308854342, + "weighted_rkd_loss": 0.005928788520395756 + }, + { + "epoch": 5.242881072026801, + "flow_matching_loss": 0.015614958759397269, + "grad_norm": 0.22036311030387878, + "learning_rate": 1.0091991923559348e-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": 0.00548774020280689, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0024738425039686263, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.013449322991073131, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.015614958759397269, + "weighted_rkd_loss": 0.006724661495536566 + }, + { + "epoch": 5.259631490787269, + "flow_matching_loss": 0.014919089665636421, + "grad_norm": 0.1609790325164795, + "learning_rate": 9.99835268639984e-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": 0.005167843447998166, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0021391444955952466, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.012474831566214562, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.014919089665636421, + "weighted_rkd_loss": 0.006237415783107281 + }, + { + "epoch": 5.276381909547739, + "flow_matching_loss": 0.013936952222138643, + "grad_norm": 0.16832859814167023, + "learning_rate": 9.904932237496913e-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": 0.0049352750414982435, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.002076266333460808, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.011946816369891167, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.013936952222138643, + "weighted_rkd_loss": 0.0059734081849455835 + }, + { + "epoch": 5.293132328308207, + "flow_matching_loss": 0.012869645096361636, + "grad_norm": 0.17330895364284515, + "learning_rate": 9.811734663363659e-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": 0.005114914570003748, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0023217051057145, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.012551534455269575, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.012869645096361636, + "weighted_rkd_loss": 0.0062757672276347876 + }, + { + "epoch": 5.309882747068677, + "flow_matching_loss": 0.01596754267811775, + "grad_norm": 0.20717377960681915, + "learning_rate": 9.718764040763875e-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": 0.0068621003534644846, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.002769121224991977, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.016493321862071753, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.01596754267811775, + "weighted_rkd_loss": 0.008246660931035876 + }, + { + "epoch": 5.326633165829146, + "flow_matching_loss": 0.013056589663028717, + "grad_norm": 0.18150459229946136, + "learning_rate": 9.626024436533777e-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": 0.005945549625903368, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.002427968825213611, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.014319068286567926, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.013056589663028717, + "weighted_rkd_loss": 0.007159534143283963 + }, + { + "epoch": 5.343383584589615, + "flow_matching_loss": 0.01568213803693652, + "grad_norm": 0.1437157690525055, + "learning_rate": 9.533519907404095e-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": 0.0047857168596237894, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.002092471101786941, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.011663904972374439, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.01568213803693652, + "weighted_rkd_loss": 0.0058319524861872194 + }, + { + "epoch": 5.360134003350084, + "flow_matching_loss": 0.017202568333595993, + "grad_norm": 0.1709163784980774, + "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": 0.00488214164506644, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0021265184856019913, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.011890801694244146, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.017202568333595993, + "weighted_rkd_loss": 0.005945400847122073 + }, + { + "epoch": 5.376884422110553, + "flow_matching_loss": 0.01590415444225073, + "grad_norm": 0.1625984013080597, + "learning_rate": 9.349232249777124e-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": 0.005498569365590811, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0023033414385281504, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01330048032104969, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.01590415444225073, + "weighted_rkd_loss": 0.006650240160524845 + }, + { + "epoch": 5.393634840871022, + "flow_matching_loss": 0.0181461731903255, + "grad_norm": 0.19519178569316864, + "learning_rate": 9.257457182618985e-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": 0.005043261009268463, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.002520814340095967, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.012607336603105068, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.0181461731903255, + "weighted_rkd_loss": 0.006303668301552534 + }, + { + "epoch": 5.410385259631491, + "flow_matching_loss": 0.014780774526298047, + "grad_norm": 0.18064932525157928, + "learning_rate": 9.165933312886944e-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": 0.006047557084821164, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0027511882479302587, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.014846302475780249, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.014780774526298047, + "weighted_rkd_loss": 0.0074231512378901245 + }, + { + "epoch": 5.42713567839196, + "flow_matching_loss": 0.013527859747409821, + "grad_norm": 0.15953907370567322, + "learning_rate": 9.07466464413155e-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": 0.004121492733247578, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0020615094457753004, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.010304494947195052, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.013527859747409821, + "weighted_rkd_loss": 0.005152247473597526 + }, + { + "epoch": 5.443886097152429, + "flow_matching_loss": 0.016063749976456165, + "grad_norm": 0.24305878579616547, + "learning_rate": 8.983655168740048e-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": 0.0049547473667189475, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0020637805806472897, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01197327538393438, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.016063749976456165, + "weighted_rkd_loss": 0.00598663769196719 + }, + { + "epoch": 5.460636515912898, + "flow_matching_loss": 0.013650599587708711, + "grad_norm": 0.1641220897436142, + "learning_rate": 8.892908867761718e-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": 0.004460714524611831, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.00191993199987337, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.010841361153870821, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.013650599587708711, + "weighted_rkd_loss": 0.005420680576935411 + }, + { + "epoch": 5.477386934673367, + "flow_matching_loss": 0.016264970973134042, + "grad_norm": 0.19726066291332245, + "learning_rate": 8.802429710733745e-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": 0.004725346667692065, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.002034914761316031, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.011485608201473952, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.016264970973134042, + "weighted_rkd_loss": 0.005742804100736976 + }, + { + "epoch": 5.494137353433836, + "flow_matching_loss": 0.014542607683688402, + "grad_norm": 0.14502941071987152, + "learning_rate": 8.71222165550757e-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": 0.005226415698416531, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0020495078642852604, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.012502339389175176, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.014542607683688402, + "weighted_rkd_loss": 0.006251169694587588 + }, + { + "epoch": 5.510887772194305, + "flow_matching_loss": 0.015074159018695354, + "grad_norm": 0.1566695272922516, + "learning_rate": 8.622288648075748e-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": 0.004941701982170343, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.00232604939956218, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.012209453340619802, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.015074159018695354, + "weighted_rkd_loss": 0.006104726670309901 + }, + { + "epoch": 5.527638190954773, + "flow_matching_loss": 0.01538830967620015, + "grad_norm": 0.1672288030385971, + "learning_rate": 8.532634622399391e-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": 0.0053394132293760775, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0020577054703608154, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.012736531719565392, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.01538830967620015, + "weighted_rkd_loss": 0.006368265859782696 + }, + { + "epoch": 5.544388609715243, + "flow_matching_loss": 0.012408527079969645, + "grad_norm": 0.13449180126190186, + "learning_rate": 8.44326350023601e-06, + "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": 0.004540071333758533, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0017385528131853788, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.010818695602938532, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.012408527079969645, + "weighted_rkd_loss": 0.005409347801469266 + }, + { + "epoch": 5.561139028475711, + "flow_matching_loss": 0.014972069207578898, + "grad_norm": 0.1624370515346527, + "learning_rate": 8.354179190968002e-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": 0.005195741541683674, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.002315680112224072, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.012707163114100695, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.014972069207578898, + "weighted_rkd_loss": 0.006353581557050347 + }, + { + "epoch": 5.577889447236181, + "flow_matching_loss": 0.017658597324043512, + "grad_norm": 0.1527981162071228, + "learning_rate": 8.265385591431661e-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": 0.005545643903315067, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0025704527739435433, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.013661740627139807, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.017658597324043512, + "weighted_rkd_loss": 0.006830870313569904 + }, + { + "epoch": 5.594639865996649, + "flow_matching_loss": 0.013222060818225145, + "grad_norm": 0.1373794972896576, + "learning_rate": 8.176886585746671e-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": 0.005418227729387581, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0022093447158113124, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.013045800291001797, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.013222060818225145, + "weighted_rkd_loss": 0.0065229001455008985 + }, + { + "epoch": 5.611390284757119, + "flow_matching_loss": 0.012797685340046883, + "grad_norm": 0.16945552825927734, + "learning_rate": 8.088686045146247e-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": 0.004318593186326325, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0023255046922713517, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.010962691204622389, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.012797685340046883, + "weighted_rkd_loss": 0.005481345602311194 + }, + { + "epoch": 5.628140703517588, + "flow_matching_loss": 0.013005319610238075, + "grad_norm": 0.1570693403482437, + "learning_rate": 8.00078782780775e-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": 0.004446164006367326, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0022311551845632495, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.011123483255505561, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.013005319610238075, + "weighted_rkd_loss": 0.005561741627752781 + }, + { + "epoch": 5.644891122278057, + "flow_matching_loss": 0.01654042676091194, + "grad_norm": 0.15214891731739044, + "learning_rate": 7.913195778683979e-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": 0.0048288710881024596, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.00192901233676821, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.011586754489690066, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.01654042676091194, + "weighted_rkd_loss": 0.005793377244845033 + }, + { + "epoch": 5.661641541038526, + "flow_matching_loss": 0.014083790592849255, + "grad_norm": 0.15469446778297424, + "learning_rate": 7.825913729334908e-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": 0.00407305343542248, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0017037070298101752, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.009849813859909773, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.014083790592849255, + "weighted_rkd_loss": 0.004924906929954886 + }, + { + "epoch": 5.678391959798995, + "flow_matching_loss": 0.016759827733039856, + "grad_norm": 0.20145344734191895, + "learning_rate": 7.73894549776013e-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": 0.003940209886059165, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0016941382200457156, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.009574558027088642, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.016759827733039856, + "weighted_rkd_loss": 0.004787279013544321 + }, + { + "epoch": 5.695142378559464, + "flow_matching_loss": 0.013244044221937656, + "grad_norm": 0.2122158706188202, + "learning_rate": 7.65229488823184e-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": 0.005202788161113858, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0020280405762605367, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.012433616816997528, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.013244044221937656, + "weighted_rkd_loss": 0.006216808408498764 + }, + { + "epoch": 5.711892797319933, + "flow_matching_loss": 0.01592979095876217, + "grad_norm": 0.18316158652305603, + "learning_rate": 7.565965691128396e-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": 0.005638341279700398, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.002074102347251028, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.013350784964859485, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.01592979095876217, + "weighted_rkd_loss": 0.0066753924824297425 + }, + { + "epoch": 5.728643216080402, + "flow_matching_loss": 0.013631716649979354, + "grad_norm": 0.16366751492023468, + "learning_rate": 7.47996168276856e-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": 0.0039018418407067656, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0014133890508674085, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.009217072697356343, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.013631716649979354, + "weighted_rkd_loss": 0.004608536348678171 + }, + { + "epoch": 5.745393634840871, + "flow_matching_loss": 0.014068759791553021, + "grad_norm": 0.14519792795181274, + "learning_rate": 7.394286625246258e-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": 0.005458051455207169, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.002536578814033419, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.013452681805938482, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.014068759791553021, + "weighted_rkd_loss": 0.006726340902969241 + }, + { + "epoch": 5.76214405360134, + "flow_matching_loss": 0.013327827304601669, + "grad_norm": 0.15125919878482819, + "learning_rate": 7.308944266266063e-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": 0.004904886498115957, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.002017999248346314, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.011827772436663508, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.013327827304601669, + "weighted_rkd_loss": 0.005913886218331754 + }, + { + "epoch": 5.778894472361809, + "flow_matching_loss": 0.01448378199711442, + "grad_norm": 0.16186784207820892, + "learning_rate": 7.223938338979233e-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": 0.004507014737464487, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0019386717118322848, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01095270118676126, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.01448378199711442, + "weighted_rkd_loss": 0.00547635059338063 + }, + { + "epoch": 5.795644891122278, + "flow_matching_loss": 0.015600308123975993, + "grad_norm": 0.1659005880355835, + "learning_rate": 7.139272561820397e-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": 0.004978228081017733, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.00196147394599393, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.011917930282652377, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.015600308123975993, + "weighted_rkd_loss": 0.005958965141326189 + }, + { + "epoch": 5.812395309882747, + "flow_matching_loss": 0.01297576930373907, + "grad_norm": 0.16415287554264069, + "learning_rate": 7.054950638344936e-06, + "loss": 0.0179, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.004142170492559672, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.001531300344504416, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.009815641259774566, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.01297576930373907, + "weighted_rkd_loss": 0.004907820629887283 + }, + { + "epoch": 5.8291457286432165, + "flow_matching_loss": 0.013271164055913687, + "grad_norm": 0.16550561785697937, + "learning_rate": 6.970976257066941e-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": 0.0046145610976964235, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0018462048145011067, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.011075327079743147, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.013271164055913687, + "weighted_rkd_loss": 0.005537663539871574 + }, + { + "epoch": 5.8458961474036855, + "flow_matching_loss": 0.015127340983599424, + "grad_norm": 0.15099075436592102, + "learning_rate": 6.887353091297896e-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": 0.004318542499095202, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0018304343800991773, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01046751942485571, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.015127340983599424, + "weighted_rkd_loss": 0.005233759712427855 + }, + { + "epoch": 5.8626465661641545, + "flow_matching_loss": 0.01355399824678898, + "grad_norm": 0.14437517523765564, + "learning_rate": 6.804084798985965e-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": 0.004264809051528573, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0019032558193430305, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01043287394568324, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.01355399824678898, + "weighted_rkd_loss": 0.00521643697284162 + }, + { + "epoch": 5.8793969849246235, + "flow_matching_loss": 0.014860629010945559, + "grad_norm": 0.1718394011259079, + "learning_rate": 6.721175022556009e-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": 0.00439409245736897, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0017511130310595036, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.010539297945797443, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.014860629010945559, + "weighted_rkd_loss": 0.005269648972898722 + }, + { + "epoch": 5.8961474036850925, + "flow_matching_loss": 0.01369728147983551, + "grad_norm": 0.20387037098407745, + "learning_rate": 6.638627388750245e-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": 0.004711744654923677, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0019726135593373328, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.0113961027469486, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.01369728147983551, + "weighted_rkd_loss": 0.0056980513734743 + }, + { + "epoch": 5.9128978224455615, + "flow_matching_loss": 0.01615075683221221, + "grad_norm": 0.1426304429769516, + "learning_rate": 6.556445508469583e-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": 0.004828235320746899, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.001961126539390534, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.011617597192525864, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.01615075683221221, + "weighted_rkd_loss": 0.005808798596262932 + }, + { + "epoch": 5.9296482412060305, + "flow_matching_loss": 0.013974253181368112, + "grad_norm": 0.14448580145835876, + "learning_rate": 6.474632976615708e-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": 0.004280661465600133, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0018652994302101433, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.010426622349768877, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.013974253181368112, + "weighted_rkd_loss": 0.0052133111748844385 + }, + { + "epoch": 5.9463986599664995, + "flow_matching_loss": 0.013339675264433026, + "grad_norm": 0.14109483361244202, + "learning_rate": 6.393193371933791e-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": 0.004555360320955515, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0015687765087932348, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.010679497197270394, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.013339675264433026, + "weighted_rkd_loss": 0.005339748598635197 + }, + { + "epoch": 5.9631490787269685, + "flow_matching_loss": 0.012223204039037228, + "grad_norm": 0.15760351717472076, + "learning_rate": 6.312130256855977e-06, + "loss": 0.0172, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.004092269181273878, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0018160081468522548, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.010000546555966138, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.012223204039037228, + "weighted_rkd_loss": 0.005000273277983069 + }, + { + "epoch": 5.9798994974874375, + "flow_matching_loss": 0.013196866121143102, + "grad_norm": 0.13676081597805023, + "learning_rate": 6.23144717734552e-06, + "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": 0.0035694713238626718, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0014613688690587878, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.008600311633199454, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.013196866121143102, + "weighted_rkd_loss": 0.004300155816599727 + }, + { + "epoch": 5.9966499162479066, + "flow_matching_loss": 0.014821585360914468, + "grad_norm": 0.23386721312999725, + "learning_rate": 6.151147662741695e-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": 0.005108627933077514, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0021549923927523195, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01237224810756743, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.014821585360914468, + "weighted_rkd_loss": 0.006186124053783715 + }, + { + "epoch": 6.013400335008376, + "flow_matching_loss": 0.015139726269990206, + "grad_norm": 0.19289110600948334, + "learning_rate": 6.071235225605415e-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": 0.004597576847299933, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0017556668259203433, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.010950820334255696, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.015139726269990206, + "weighted_rkd_loss": 0.005475410167127848 + }, + { + "epoch": 6.030150753768845, + "flow_matching_loss": 0.013879562634974718, + "grad_norm": 0.21247315406799316, + "learning_rate": 5.991713361565546e-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": 0.004605715093202889, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0019215092528611422, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01113293943926692, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.013879562634974718, + "weighted_rkd_loss": 0.00556646971963346 + }, + { + "epoch": 6.046901172529314, + "flow_matching_loss": 0.011976712848991156, + "grad_norm": 0.20955568552017212, + "learning_rate": 5.912585549166051e-06, + "loss": 0.0172, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.004337595240212977, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0017820636508986353, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.010457254108041525, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.011976712848991156, + "weighted_rkd_loss": 0.005228627054020762 + }, + { + "epoch": 6.063651591289783, + "flow_matching_loss": 0.014497782150283457, + "grad_norm": 0.14614950120449066, + "learning_rate": 5.833855249713777e-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": 0.004382647341117263, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0017091153538785874, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.010474410047754646, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.014497782150283457, + "weighted_rkd_loss": 0.005237205023877323 + }, + { + "epoch": 6.080402010050252, + "flow_matching_loss": 0.014638572558760643, + "grad_norm": 0.17804445326328278, + "learning_rate": 5.7555259071270645e-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": 0.004319529328495264, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.002127594541525468, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.010766653111204505, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.014638572558760643, + "weighted_rkd_loss": 0.0053833265556022525 + }, + { + "epoch": 6.097152428810721, + "flow_matching_loss": 0.012663161847740412, + "grad_norm": 0.14908669888973236, + "learning_rate": 5.677600947785117e-06, + "loss": 0.0171, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.0037628420628607272, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0013517275685444475, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.00887741162441671, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.012663161847740412, + "weighted_rkd_loss": 0.004438705812208355 + }, + { + "epoch": 6.11390284757119, + "flow_matching_loss": 0.014881637273356318, + "grad_norm": 0.2055513858795166, + "learning_rate": 5.600083780378099e-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": 0.004984625265933573, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.002195600490085781, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.012164850998669863, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.014881637273356318, + "weighted_rkd_loss": 0.006082425499334932 + }, + { + "epoch": 6.130653266331659, + "flow_matching_loss": 0.014626818615943194, + "grad_norm": 0.18061131238937378, + "learning_rate": 5.522977795758038e-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": 0.004626819328404963, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.001866737351519987, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.011120375851169229, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.014626818615943194, + "weighted_rkd_loss": 0.0055601879255846145 + }, + { + "epoch": 6.147403685092128, + "flow_matching_loss": 0.013677647057920694, + "grad_norm": 0.1852584183216095, + "learning_rate": 5.4462863667904805e-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": 0.00379023642744869, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0017943391227163374, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.0093748118262738, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.013677647057920694, + "weighted_rkd_loss": 0.0046874059131369 + }, + { + "epoch": 6.164154103852597, + "flow_matching_loss": 0.013752449536696076, + "grad_norm": 0.15452587604522705, + "learning_rate": 5.370012848206983e-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": 0.003933401824906468, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0015195835614576936, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.009386387234553695, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.013752449536696076, + "weighted_rkd_loss": 0.004693193617276848 + }, + { + "epoch": 6.180904522613066, + "flow_matching_loss": 0.013424647925421595, + "grad_norm": 0.16993391513824463, + "learning_rate": 5.29416057645834e-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": 0.004384127364028245, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0018952976912260055, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01066355244256556, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.013424647925421595, + "weighted_rkd_loss": 0.00533177622128278 + }, + { + "epoch": 6.197654941373535, + "flow_matching_loss": 0.014398802910000086, + "grad_norm": 0.15566317737102509, + "learning_rate": 5.218732869568635e-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": 0.005112427612766623, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0022814268944785, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01250628214329481, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.014398802910000086, + "weighted_rkd_loss": 0.006253141071647405 + }, + { + "epoch": 6.214405360134004, + "flow_matching_loss": 0.013579555507749318, + "grad_norm": 0.15493236482143402, + "learning_rate": 5.14373302699013e-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": 0.004654577840119601, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0017654628201853484, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.011074618343263865, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.013579555507749318, + "weighted_rkd_loss": 0.005537309171631932 + }, + { + "epoch": 6.231155778894473, + "flow_matching_loss": 0.012991676107048988, + "grad_norm": 0.14714433252811432, + "learning_rate": 5.069164329458894e-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": 0.004408569377847016, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0017401628778316081, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01055730152875185, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.012991676107048988, + "weighted_rkd_loss": 0.005278650764375925 + }, + { + "epoch": 6.247906197654942, + "flow_matching_loss": 0.014295129012316466, + "grad_norm": 0.17364512383937836, + "learning_rate": 4.995030038851342e-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": 0.004136458213906735, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0018542096950113773, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.010127126052975655, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.014295129012316466, + "weighted_rkd_loss": 0.005063563026487828 + }, + { + "epoch": 6.264656616415411, + "flow_matching_loss": 0.01341222645714879, + "grad_norm": 0.20091146230697632, + "learning_rate": 4.9213333980415e-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": 0.0047332413727417585, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0018039730959571898, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.011270455876365304, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.01341222645714879, + "weighted_rkd_loss": 0.005635227938182652 + }, + { + "epoch": 6.28140703517588, + "flow_matching_loss": 0.014535247907042503, + "grad_norm": 0.14015261828899384, + "learning_rate": 4.848077630759175e-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": 0.004209902766160667, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.00176135414512828, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01018115975894034, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.014535247907042503, + "weighted_rkd_loss": 0.00509057987947017 + }, + { + "epoch": 6.298157453936349, + "flow_matching_loss": 0.013406949071213602, + "grad_norm": 0.16716063022613525, + "learning_rate": 4.775265941448951e-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": 0.003995424625463784, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0015201234142296015, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.00951097272336483, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.013406949071213602, + "weighted_rkd_loss": 0.004755486361682415 + }, + { + "epoch": 6.314907872696818, + "flow_matching_loss": 0.013139669876545667, + "grad_norm": 0.14437201619148254, + "learning_rate": 4.702901515129989e-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": 0.0044357549166306855, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0017475514672696591, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.010619061347097159, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.013139669876545667, + "weighted_rkd_loss": 0.0053095306735485796 + }, + { + "epoch": 6.331658291457287, + "flow_matching_loss": 0.013105375785380602, + "grad_norm": 0.13337287306785583, + "learning_rate": 4.630987517256734e-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": 0.00446128414478153, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.00155455416534096, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.010477122571319341, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.013105375785380602, + "weighted_rkd_loss": 0.0052385612856596705 + }, + { + "epoch": 6.348408710217756, + "flow_matching_loss": 0.01332718194462359, + "grad_norm": 0.15027029812335968, + "learning_rate": 4.559527093580412e-06, + "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": 0.004570346092805266, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.002078494516899809, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.011219186754897238, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.01332718194462359, + "weighted_rkd_loss": 0.005609593377448619 + }, + { + "epoch": 6.365159128978225, + "flow_matching_loss": 0.0158365941606462, + "grad_norm": 0.15022490918636322, + "learning_rate": 4.488523370011456e-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": 0.004662989941425622, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0019907014677301047, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.011316681420430541, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.0158365941606462, + "weighted_rkd_loss": 0.005658340710215271 + }, + { + "epoch": 6.381909547738694, + "flow_matching_loss": 0.012740066088736058, + "grad_norm": 0.21718840301036835, + "learning_rate": 4.417979452482766e-06, + "loss": 0.0179, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.00425658579915762, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0017564637295436115, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01026963540352881, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.012740066088736058, + "weighted_rkd_loss": 0.005134817701764405 + }, + { + "epoch": 6.398659966499163, + "flow_matching_loss": 0.012215562909841538, + "grad_norm": 0.1450190246105194, + "learning_rate": 4.347898426813811e-06, + "loss": 0.0179, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.004763328027911485, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0018555120564997195, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.011382168158888817, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.012215562909841538, + "weighted_rkd_loss": 0.005691084079444408 + }, + { + "epoch": 6.415410385259632, + "flow_matching_loss": 0.01273781326599419, + "grad_norm": 0.14332900941371918, + "learning_rate": 4.278283358575694e-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": 0.0041654838249087335, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0016894752159714698, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01002044277265668, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.01273781326599419, + "weighted_rkd_loss": 0.00501022138632834 + }, + { + "epoch": 6.432160804020101, + "flow_matching_loss": 0.013954770937561988, + "grad_norm": 0.16812093555927277, + "learning_rate": 4.2091372929570145e-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": 0.003278344660066068, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0015723046730272473, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.008128994004800915, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.013954770937561988, + "weighted_rkd_loss": 0.004064497002400458 + }, + { + "epoch": 6.44891122278057, + "flow_matching_loss": 0.015501107182353735, + "grad_norm": 0.16985221207141876, + "learning_rate": 4.1404632546306934e-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": 0.004027113923802972, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0016444221953861415, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.009698650008067489, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.015501107182353735, + "weighted_rkd_loss": 0.0048493250040337445 + }, + { + "epoch": 6.465661641541039, + "flow_matching_loss": 0.014016295969486236, + "grad_norm": 0.16254419088363647, + "learning_rate": 4.0722642476216286e-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": 0.0042627822840586305, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0018491683877073228, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.010374732967466115, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.014016295969486236, + "weighted_rkd_loss": 0.005187366483733058 + }, + { + "epoch": 6.482412060301508, + "flow_matching_loss": 0.013069759542122483, + "grad_norm": 0.1539461463689804, + "learning_rate": 4.00454325517533e-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": 0.004568273038603366, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.002030610025394708, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.011167156137526035, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.013069759542122483, + "weighted_rkd_loss": 0.0055835780687630175 + }, + { + "epoch": 6.499162479061977, + "flow_matching_loss": 0.013110130187124014, + "grad_norm": 0.1754523068666458, + "learning_rate": 3.937303239627393e-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": 0.004449385032057762, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0018155326717533172, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.010714302537962794, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.013110130187124014, + "weighted_rkd_loss": 0.005357151268981397 + }, + { + "epoch": 6.515912897822446, + "flow_matching_loss": 0.014289885107427836, + "grad_norm": 0.13485778868198395, + "learning_rate": 3.870547142273918e-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": 0.0033146835397928954, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0012253299821168185, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.007854697108268738, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.014289885107427836, + "weighted_rkd_loss": 0.003927348554134369 + }, + { + "epoch": 6.532663316582915, + "flow_matching_loss": 0.013430317863821983, + "grad_norm": 0.13589271903038025, + "learning_rate": 3.8042778832428736e-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": 0.004406502214260399, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0016344427305739373, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.010447447188198566, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.013430317863821983, + "weighted_rkd_loss": 0.005223723594099283 + }, + { + "epoch": 6.549413735343384, + "flow_matching_loss": 0.0178004146553576, + "grad_norm": 0.16389556229114532, + "learning_rate": 3.7384983613663195e-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": 0.0039194801589474085, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0016993709024973214, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.009538331348448992, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.0178004146553576, + "weighted_rkd_loss": 0.004769165674224496 + }, + { + "epoch": 6.566164154103853, + "flow_matching_loss": 0.014947613794356584, + "grad_norm": 0.202504500746727, + "learning_rate": 3.6732114540536503e-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": 0.004656322370283306, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.001776739489287138, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.011089384276419877, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.014947613794356584, + "weighted_rkd_loss": 0.005544692138209939 + }, + { + "epoch": 6.582914572864322, + "flow_matching_loss": 0.015553992334753275, + "grad_norm": 0.19275301694869995, + "learning_rate": 3.608420017165683e-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": 0.004351094085723162, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0019604666507802905, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.010662654880434275, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.015553992334753275, + "weighted_rkd_loss": 0.005331327440217137 + }, + { + "epoch": 6.599664991624791, + "flow_matching_loss": 0.013738202583044767, + "grad_norm": 0.16886192560195923, + "learning_rate": 3.5441268848897664e-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": 0.004141055117361248, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0016448502079583704, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.009926960431039333, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.013738202583044767, + "weighted_rkd_loss": 0.0049634802155196665 + }, + { + "epoch": 6.61641541038526, + "flow_matching_loss": 0.013459889078512787, + "grad_norm": 0.11751831322908401, + "learning_rate": 3.480334869615792e-06, + "loss": 0.0179, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.0036029696464538576, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0016099985863547773, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.008815937908366323, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.013459889078512787, + "weighted_rkd_loss": 0.004407968954183161 + }, + { + "epoch": 6.633165829145729, + "flow_matching_loss": 0.012848709896206855, + "grad_norm": 0.18301798403263092, + "learning_rate": 3.417046761813155e-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": 0.003821439016610384, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.001900578138884157, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.009543456276878715, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.012848709896206855, + "weighted_rkd_loss": 0.0047717281384393575 + }, + { + "epoch": 6.649916247906198, + "flow_matching_loss": 0.01411162167787552, + "grad_norm": 0.1493939906358719, + "learning_rate": 3.3542653299087263e-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": 0.003471779148094356, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0017120651376899333, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.008655623439699412, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.01411162167787552, + "weighted_rkd_loss": 0.004327811719849706 + }, + { + "epoch": 6.666666666666667, + "flow_matching_loss": 0.010548923909664155, + "grad_norm": 0.14926926791667938, + "learning_rate": 3.2919933201657117e-06, + "loss": 0.0156, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.004148266627453268, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.001756682724226266, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.010053216014057398, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.010548923909664155, + "weighted_rkd_loss": 0.005026608007028699 + }, + { + "epoch": 6.683417085427136, + "flow_matching_loss": 0.014358358830213547, + "grad_norm": 0.1853066384792328, + "learning_rate": 3.2302334565635584e-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": 0.003923210455104709, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.001574484136654064, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.009420905215665697, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.014358358830213547, + "weighted_rkd_loss": 0.004710452607832849 + }, + { + "epoch": 6.700167504187605, + "flow_matching_loss": 0.01270328126847744, + "grad_norm": 0.19508641958236694, + "learning_rate": 3.1689884406787674e-06, + "loss": 0.0172, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.003606516565196216, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0017539363470859826, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.008966969326138496, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.01270328126847744, + "weighted_rkd_loss": 0.004483484663069248 + }, + { + "epoch": 6.716917922948074, + "flow_matching_loss": 0.01392309032380581, + "grad_norm": 0.16966725885868073, + "learning_rate": 3.1082609515667424e-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": 0.0044722165912389755, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0017818786611314863, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.010726311895996331, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.01392309032380581, + "weighted_rkd_loss": 0.005363155947998166 + }, + { + "epoch": 6.733668341708543, + "flow_matching_loss": 0.012681249156594276, + "grad_norm": 0.20630866289138794, + "learning_rate": 3.048053645644594e-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": 0.004192276950925588, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0016123710433021189, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.009996925015002489, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.012681249156594276, + "weighted_rkd_loss": 0.0049984625075012445 + }, + { + "epoch": 6.750418760469012, + "flow_matching_loss": 0.013639771938323974, + "grad_norm": 0.16677002608776093, + "learning_rate": 2.988369156574916e-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": 0.0040628035785630345, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0015570100396871567, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.00968261705711484, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.013639771938323974, + "weighted_rkd_loss": 0.00484130852855742 + }, + { + "epoch": 6.767169179229481, + "flow_matching_loss": 0.01363765113055706, + "grad_norm": 0.12832656502723694, + "learning_rate": 2.9292100951506225e-06, + "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": 0.004342612088657915, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0018042451236397028, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.010489469300955534, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.01363765113055706, + "weighted_rkd_loss": 0.005244734650477767 + }, + { + "epoch": 6.78391959798995, + "flow_matching_loss": 0.013352138735353947, + "grad_norm": 0.15868757665157318, + "learning_rate": 2.8705790491807033e-06, + "loss": 0.0179, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.003720019944012165, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0015638376004062592, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.009003877593204378, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.013352138735353947, + "weighted_rkd_loss": 0.004501938796602189 + }, + { + "epoch": 6.800670016750419, + "flow_matching_loss": 0.012715986743569375, + "grad_norm": 0.18388469517230988, + "learning_rate": 2.8124785833770433e-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": 0.004414536338299513, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0018238211283460259, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01065289373509586, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.012715986743569375, + "weighted_rkd_loss": 0.00532644686754793 + }, + { + "epoch": 6.817420435510888, + "flow_matching_loss": 0.014656269457191229, + "grad_norm": 0.20179112255573273, + "learning_rate": 2.754911239242241e-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": 0.004632821911945939, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0020091856247745456, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.011274829413741827, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.014656269457191229, + "weighted_rkd_loss": 0.005637414706870914 + }, + { + "epoch": 6.834170854271357, + "flow_matching_loss": 0.012329512694850563, + "grad_norm": 0.19481857120990753, + "learning_rate": 2.6978795349584204e-06, + "loss": 0.0179, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.004721343913115561, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.001782020868267864, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01122470865957439, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.012329512694850563, + "weighted_rkd_loss": 0.005612354329787195 + }, + { + "epoch": 6.850921273031826, + "flow_matching_loss": 0.013932113721966743, + "grad_norm": 0.17180730402469635, + "learning_rate": 2.641385965277088e-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": 0.004377597151324153, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0019918588106520472, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.010747053101658821, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.013932113721966743, + "weighted_rkd_loss": 0.005373526550829411 + }, + { + "epoch": 6.867671691792295, + "flow_matching_loss": 0.012642812356352806, + "grad_norm": 0.17917793989181519, + "learning_rate": 2.5854330014099923e-06, + "loss": 0.017, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.0035743531887419524, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0015058233228046447, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.008654529694467783, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.012642812356352806, + "weighted_rkd_loss": 0.0043272648472338915 + }, + { + "epoch": 6.884422110552764, + "flow_matching_loss": 0.01389442621730268, + "grad_norm": 0.12711286544799805, + "learning_rate": 2.5300230909210298e-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": 0.0035649200435727835, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0015694846806582063, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.008699324633926152, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.01389442621730268, + "weighted_rkd_loss": 0.004349662316963076 + }, + { + "epoch": 6.901172529313233, + "flow_matching_loss": 0.012807622458785773, + "grad_norm": 0.19579429924488068, + "learning_rate": 2.4751586576191983e-06, + "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": 0.004366207169368863, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0018814566778019071, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01061387094669044, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.012807622458785773, + "weighted_rkd_loss": 0.00530693547334522 + }, + { + "epoch": 6.917922948073702, + "flow_matching_loss": 0.012661585444584489, + "grad_norm": 0.1477232426404953, + "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": 0.004671722766943276, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0016743403277359903, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01101778601296246, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.012661585444584489, + "weighted_rkd_loss": 0.00550889300648123 + }, + { + "epoch": 6.934673366834171, + "flow_matching_loss": 0.012656646966934203, + "grad_norm": 0.17730483412742615, + "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": 0.004852761048823595, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0018881546217016875, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.011593676824122667, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.012656646966934203, + "weighted_rkd_loss": 0.005796838412061333 + }, + { + "epoch": 6.95142378559464, + "flow_matching_loss": 0.011845463747158646, + "grad_norm": 0.1716385930776596, + "learning_rate": 2.313862100383404e-06, + "loss": 0.0164, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.0035892325220629575, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0018717827624641359, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.009050247818231582, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.011845463747158646, + "weighted_rkd_loss": 0.004525123909115791 + }, + { + "epoch": 6.968174204355109, + "flow_matching_loss": 0.012647939100861549, + "grad_norm": 0.1931212693452835, + "learning_rate": 2.2612033351327406e-06, + "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": 0.004044805024750531, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0015366453211754561, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.009626255370676517, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.012647939100861549, + "weighted_rkd_loss": 0.004813127685338259 + }, + { + "epoch": 6.984924623115578, + "flow_matching_loss": 0.01442919559776783, + "grad_norm": 0.19145072996616364, + "learning_rate": 2.2091018061161788e-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": 0.0043393786996603016, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.002081942302174866, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.010760699631646276, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.01442919559776783, + "weighted_rkd_loss": 0.005380349815823138 + }, + { + "epoch": 7.001675041876047, + "flow_matching_loss": 0.013325535040348769, + "grad_norm": 0.16990314424037933, + "learning_rate": 2.15755979242338e-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": 0.00459274286404252, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.00216435887850821, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.011349844746291637, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.013325535040348769, + "weighted_rkd_loss": 0.005674922373145819 + }, + { + "epoch": 7.018425460636516, + "flow_matching_loss": 0.015028421999886632, + "grad_norm": 0.1645367592573166, + "learning_rate": 2.106579548669e-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": 0.00421563507989049, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.001719225279521197, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.010150495450943708, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.015028421999886632, + "weighted_rkd_loss": 0.005075247725471854 + }, + { + "epoch": 7.035175879396985, + "flow_matching_loss": 0.011063610622659325, + "grad_norm": 0.14495235681533813, + "learning_rate": 2.0561633048940477e-06, + "loss": 0.0154, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.0035904986783862115, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0014928206335753202, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.008673817990347743, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.011063610622659325, + "weighted_rkd_loss": 0.004336908995173872 + }, + { + "epoch": 7.051926298157454, + "flow_matching_loss": 0.014337676949799061, + "grad_norm": 0.14312154054641724, + "learning_rate": 2.006313266468355e-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": 0.0042888281866908075, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0018349151243455708, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.010412571532651783, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.014337676949799061, + "weighted_rkd_loss": 0.005206285766325891 + }, + { + "epoch": 7.068676716917923, + "flow_matching_loss": 0.011880451068282127, + "grad_norm": 0.17340722680091858, + "learning_rate": 1.957031613994097e-06, + "loss": 0.0163, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.0037391659570857882, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0014357647683937102, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.008914096700027584, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.011880451068282127, + "weighted_rkd_loss": 0.004457048350013792 + }, + { + "epoch": 7.085427135678392, + "flow_matching_loss": 0.013384684594348073, + "grad_norm": 0.14712361991405487, + "learning_rate": 1.9083205032103995e-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": 0.0043742188951000575, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0020250638597644867, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01077350163832307, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.013384684594348073, + "weighted_rkd_loss": 0.005386750819161535 + }, + { + "epoch": 7.102177554438861, + "flow_matching_loss": 0.014141054265201092, + "grad_norm": 0.14053156971931458, + "learning_rate": 1.8601820648990536e-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": 0.003534198086708784, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0016627753793727607, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.008731171535328031, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.014141054265201092, + "weighted_rkd_loss": 0.004365585767664016 + }, + { + "epoch": 7.11892797319933, + "flow_matching_loss": 0.012220955546945333, + "grad_norm": 0.14366623759269714, + "learning_rate": 1.8126184047912936e-06, + "loss": 0.0174, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.0042095793643966315, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.002016812388319522, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.010435971105471254, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.012220955546945333, + "weighted_rkd_loss": 0.005217985552735627 + }, + { + "epoch": 7.135678391959799, + "flow_matching_loss": 0.012571109691634774, + "grad_norm": 0.14746060967445374, + "learning_rate": 1.7656316034757024e-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": 0.004236602829769254, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0017786037409678102, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.010251809330657124, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.012571109691634774, + "weighted_rkd_loss": 0.005125904665328562 + }, + { + "epoch": 7.152428810720268, + "flow_matching_loss": 0.013217747118324042, + "grad_norm": 0.16644246876239777, + "learning_rate": 1.7192237163071816e-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": 0.004921759618446231, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0020453046192415057, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.011888823937624693, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.013217747118324042, + "weighted_rkd_loss": 0.0059444119688123465 + }, + { + "epoch": 7.169179229480737, + "flow_matching_loss": 0.013593366229906678, + "grad_norm": 0.1624123603105545, + "learning_rate": 1.6733967733170573e-06, + "loss": 0.0179, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.00354135453235358, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0015589617833029479, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.008641670923680067, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.013593366229906678, + "weighted_rkd_loss": 0.0043208354618400335 + }, + { + "epoch": 7.185929648241206, + "flow_matching_loss": 0.011336163431406022, + "grad_norm": 0.15716801583766937, + "learning_rate": 1.628152779124279e-06, + "loss": 0.0166, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.004345083329826594, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.001911070611095056, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.010601237136870622, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.011336163431406022, + "weighted_rkd_loss": 0.005300618568435311 + }, + { + "epoch": 7.202680067001675, + "flow_matching_loss": 0.012439127638936042, + "grad_norm": 0.1175704151391983, + "learning_rate": 1.583493712847715e-06, + "loss": 0.0172, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.003965030517429113, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0015500714420340956, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.009480132395401597, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.012439127638936042, + "weighted_rkd_loss": 0.004740066197700799 + }, + { + "epoch": 7.219430485762144, + "flow_matching_loss": 0.01372097311541438, + "grad_norm": 0.14474470913410187, + "learning_rate": 1.5394215280196084e-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": 0.004111591842956841, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0015248423907905817, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.009748026169836521, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.01372097311541438, + "weighted_rkd_loss": 0.004874013084918261 + }, + { + "epoch": 7.236180904522613, + "flow_matching_loss": 0.012772003002464772, + "grad_norm": 0.15785634517669678, + "learning_rate": 1.4959381525000904e-06, + "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": 0.004256247635930776, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.001629356446210295, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.010141851706430316, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.012772003002464772, + "weighted_rkd_loss": 0.005070925853215158 + }, + { + "epoch": 7.252931323283082, + "flow_matching_loss": 0.014338200632482766, + "grad_norm": 0.160471573472023, + "learning_rate": 1.453045488392879e-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": 0.004117710143327713, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0016662140609696508, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.009901634324342013, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.014338200632482766, + "weighted_rkd_loss": 0.004950817162171006 + }, + { + "epoch": 7.269681742043551, + "flow_matching_loss": 0.012580685364082456, + "grad_norm": 0.13657648861408234, + "learning_rate": 1.4107454119620484e-06, + "loss": 0.0169, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.0036080827238038184, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0013286747212987392, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.008544840198010207, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.012580685364082456, + "weighted_rkd_loss": 0.0042724200990051035 + }, + { + "epoch": 7.28643216080402, + "flow_matching_loss": 0.013763728784397244, + "grad_norm": 0.16480371356010437, + "learning_rate": 1.3690397735499778e-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": 0.004622260120231658, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0022792898351326587, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.011523810215294361, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.013763728784397244, + "weighted_rkd_loss": 0.0057619051076471806 + }, + { + "epoch": 7.303182579564489, + "flow_matching_loss": 0.014634063653647899, + "grad_norm": 0.1543356031179428, + "learning_rate": 1.327930397496403e-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": 0.004943400342017412, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0021274060709401965, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01201420659199357, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.014634063653647899, + "weighted_rkd_loss": 0.006007103295996785 + }, + { + "epoch": 7.319932998324958, + "flow_matching_loss": 0.011182602029293776, + "grad_norm": 0.23041939735412598, + "learning_rate": 1.287419082058604e-06, + "loss": 0.0161, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.003948941268026829, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0018433966557495295, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.009741279110312463, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.011182602029293776, + "weighted_rkd_loss": 0.004870639555156231 + }, + { + "epoch": 7.336683417085427, + "flow_matching_loss": 0.012292292807251215, + "grad_norm": 0.1963876187801361, + "learning_rate": 1.2475075993327589e-06, + "loss": 0.0169, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.003892730991356075, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.001485452667111531, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.009270914644002915, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.012292292807251215, + "weighted_rkd_loss": 0.0046354573220014576 + }, + { + "epoch": 7.353433835845896, + "flow_matching_loss": 0.014200325449928641, + "grad_norm": 0.19687843322753906, + "learning_rate": 1.208197695176413e-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": 0.00432422065641731, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.002694979298394173, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01134342052973807, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.014200325449928641, + "weighted_rkd_loss": 0.005671710264869035 + }, + { + "epoch": 7.370184254606365, + "flow_matching_loss": 0.011306853778660298, + "grad_norm": 0.16218173503875732, + "learning_rate": 1.1694910891321259e-06, + "loss": 0.0164, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.004136376339010895, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.001833427872043103, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.010106180747970938, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.011306853778660298, + "weighted_rkd_loss": 0.005053090373985469 + }, + { + "epoch": 7.386934673366834, + "flow_matching_loss": 0.013065098226070404, + "grad_norm": 0.1709485948085785, + "learning_rate": 1.1313894743522363e-06, + "loss": 0.0174, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.0036178565234877167, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0015012272633612156, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.008736940333619714, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.013065098226070404, + "weighted_rkd_loss": 0.004368470166809857 + }, + { + "epoch": 7.403685092127303, + "flow_matching_loss": 0.01347271353006363, + "grad_norm": 0.1800774782896042, + "learning_rate": 1.0938945175248016e-06, + "loss": 0.0179, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.003690591617487371, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0015590842347592115, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.00894026760943234, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.01347271353006363, + "weighted_rkd_loss": 0.00447013380471617 + }, + { + "epoch": 7.420435510887772, + "flow_matching_loss": 0.013602646300569177, + "grad_norm": 0.1115361675620079, + "learning_rate": 1.0570078588007086e-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": 0.0041832276852801446, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0017717199341859668, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.010138175310567021, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.013602646300569177, + "weighted_rkd_loss": 0.0050690876552835105 + }, + { + "epoch": 7.437185929648241, + "flow_matching_loss": 0.011779169458895922, + "grad_norm": 0.13396503031253815, + "learning_rate": 1.0207311117219009e-06, + "loss": 0.0168, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.004144427995197475, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.001768887892831117, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.010057743871584535, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.011779169458895922, + "weighted_rkd_loss": 0.005028871935792267 + }, + { + "epoch": 7.45393634840871, + "flow_matching_loss": 0.011645021988078952, + "grad_norm": 0.11680609732866287, + "learning_rate": 9.850658631508198e-07, + "loss": 0.0161, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.003669659234583378, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0015738890622742474, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.008913207612931729, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.011645021988078952, + "weighted_rkd_loss": 0.0044566038064658645 + }, + { + "epoch": 7.4706867671691795, + "flow_matching_loss": 0.012487611640244723, + "grad_norm": 0.15301018953323364, + "learning_rate": 9.500136732009751e-07, + "loss": 0.0169, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.0036924283020198347, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.001380647497717291, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.008765504136681557, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.012487611640244723, + "weighted_rkd_loss": 0.0043827520683407785 + }, + { + "epoch": 7.4874371859296485, + "flow_matching_loss": 0.01187393870204687, + "grad_norm": 0.13816434144973755, + "learning_rate": 9.155760751687054e-07, + "loss": 0.0162, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.0035726704518310726, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0014808060775976628, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.008626147080212832, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.01187393870204687, + "weighted_rkd_loss": 0.004313073540106416 + }, + { + "epoch": 7.5041876046901175, + "flow_matching_loss": 0.011787522397935391, + "grad_norm": 0.1930665522813797, + "learning_rate": 8.817545754661155e-07, + "loss": 0.0165, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.0038458931958302855, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0016517811745870858, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.009343567583709956, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.011787522397935391, + "weighted_rkd_loss": 0.004671783791854978 + }, + { + "epoch": 7.5209380234505865, + "flow_matching_loss": 0.014746030792593956, + "grad_norm": 0.15736708045005798, + "learning_rate": 8.485506535551668e-07, + "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": 0.004147592210210859, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0017917696095537394, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.010086953965947032, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.014746030792593956, + "weighted_rkd_loss": 0.005043476982973516 + }, + { + "epoch": 7.5376884422110555, + "flow_matching_loss": 0.011041402723640204, + "grad_norm": 0.15687477588653564, + "learning_rate": 8.159657618829685e-07, + "loss": 0.0155, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.003685565129853785, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0015495843777898698, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.008920714585110545, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.011041402723640204, + "weighted_rkd_loss": 0.004460357292555272 + }, + { + "epoch": 7.5544388609715245, + "flow_matching_loss": 0.013714790437370538, + "grad_norm": 0.12664705514907837, + "learning_rate": 7.840013258182471e-07, + "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": 0.0038969279499724506, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0014690222684293984, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.0092628781683743, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.013714790437370538, + "weighted_rkd_loss": 0.00463143908418715 + }, + { + "epoch": 7.5711892797319935, + "flow_matching_loss": 0.014138914924114943, + "grad_norm": 0.12797820568084717, + "learning_rate": 7.526587435889809e-07, + "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": 0.0034030490089207886, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0013546278292778879, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.008160725887864827, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.014138914924114943, + "weighted_rkd_loss": 0.004080362943932414 + }, + { + "epoch": 7.5879396984924625, + "flow_matching_loss": 0.012961727753281593, + "grad_norm": 0.15495085716247559, + "learning_rate": 7.219393862212593e-07, + "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": 0.003796363272704184, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.001647556375246495, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.009240282978862525, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.012961727753281593, + "weighted_rkd_loss": 0.004620141489431262 + }, + { + "epoch": 7.6046901172529315, + "flow_matching_loss": 0.01382149439305067, + "grad_norm": 0.17086490988731384, + "learning_rate": 6.918445974792819e-07, + "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": 0.003138877125456929, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0015061549376696349, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.007783909235149622, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.01382149439305067, + "weighted_rkd_loss": 0.003891954617574811 + }, + { + "epoch": 7.6214405360134005, + "flow_matching_loss": 0.01289975387044251, + "grad_norm": 0.10146980732679367, + "learning_rate": 6.6237569380661e-07, + "loss": 0.017, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.0034003766952082513, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0013089508167468012, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.008109704172238708, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.01289975387044251, + "weighted_rkd_loss": 0.004054852086119354 + }, + { + "epoch": 7.63819095477387, + "flow_matching_loss": 0.013388771284371614, + "grad_norm": 0.19025149941444397, + "learning_rate": 6.335339642685495e-07, + "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": 0.0039990865625441074, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.00168166448129341, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.009679837571457028, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.013388771284371614, + "weighted_rkd_loss": 0.004839918785728514 + }, + { + "epoch": 7.654941373534339, + "flow_matching_loss": 0.013220328884199262, + "grad_norm": 0.1995600312948227, + "learning_rate": 6.05320670495787e-07, + "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": 0.0038794698659330605, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0017690667242277413, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.009528006473556162, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.013220328884199262, + "weighted_rkd_loss": 0.004764003236778081 + }, + { + "epoch": 7.671691792294808, + "flow_matching_loss": 0.012908745370805264, + "grad_norm": 0.18338993191719055, + "learning_rate": 5.777370466291954e-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": 0.004892875347286463, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.002224490814842284, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.012010241625830531, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.012908745370805264, + "weighted_rkd_loss": 0.0060051208129152656 + }, + { + "epoch": 7.688442211055277, + "flow_matching_loss": 0.012604714278131724, + "grad_norm": 0.17824886739253998, + "learning_rate": 5.507842992658336e-07, + "loss": 0.017, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.003669789177365601, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0014661247783806174, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.008805703138932586, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.012604714278131724, + "weighted_rkd_loss": 0.004402851569466293 + }, + { + "epoch": 7.705192629815746, + "flow_matching_loss": 0.013908560667186976, + "grad_norm": 0.1996762901544571, + "learning_rate": 5.24463607406192e-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": 0.004282183572649955, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0017393660498782992, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.010303733311593533, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.013908560667186976, + "weighted_rkd_loss": 0.0051518666557967665 + }, + { + "epoch": 7.721943048576215, + "flow_matching_loss": 0.011516436655074357, + "grad_norm": 0.15033936500549316, + "learning_rate": 4.987761224025944e-07, + "loss": 0.0165, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.004200274380855263, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0015304363565519451, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.00993098528124392, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.011516436655074357, + "weighted_rkd_loss": 0.00496549264062196 + }, + { + "epoch": 7.738693467336684, + "flow_matching_loss": 0.016077415226027368, + "grad_norm": 0.15527673065662384, + "learning_rate": 4.737229679088556e-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": 0.0033825806342065333, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0015490434423554688, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.008314204728230834, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.016077415226027368, + "weighted_rkd_loss": 0.004157102364115417 + }, + { + "epoch": 7.755443886097153, + "flow_matching_loss": 0.013937278091907502, + "grad_norm": 0.15270797908306122, + "learning_rate": 4.493052398311115e-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": 0.0038180277915671468, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0016788874345365912, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.009314942918717862, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.013937278091907502, + "weighted_rkd_loss": 0.004657471459358931 + }, + { + "epoch": 7.772194304857622, + "flow_matching_loss": 0.012719677481800317, + "grad_norm": 0.160204216837883, + "learning_rate": 4.2552400627989043e-07, + "loss": 0.0173, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.0037694335216656326, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0015906929154880344, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.00912955985404551, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.012719677481800317, + "weighted_rkd_loss": 0.004564779927022755 + }, + { + "epoch": 7.788944723618091, + "flow_matching_loss": 0.012721182778477669, + "grad_norm": 0.18550850450992584, + "learning_rate": 4.023803075233923e-07, + "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": 0.004492715140804649, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0019040567451156676, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01088948706164956, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.012721182778477669, + "weighted_rkd_loss": 0.00544474353082478 + }, + { + "epoch": 7.80569514237856, + "flow_matching_loss": 0.013466277066618204, + "grad_norm": 0.14704406261444092, + "learning_rate": 3.7987515594196997e-07, + "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": 0.0033923544222489, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0013608533889055252, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.008145562140271068, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.013466277066618204, + "weighted_rkd_loss": 0.004072781070135534 + }, + { + "epoch": 7.822445561139029, + "flow_matching_loss": 0.013312026532366872, + "grad_norm": 0.21852347254753113, + "learning_rate": 3.580095359838648e-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": 0.004508821619674563, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0016108532319776714, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.010628496576100589, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.013312026532366872, + "weighted_rkd_loss": 0.005314248288050294 + }, + { + "epoch": 7.839195979899498, + "flow_matching_loss": 0.012858473043888808, + "grad_norm": 0.17716294527053833, + "learning_rate": 3.3678440412212576e-07, + "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": 0.004035146767273545, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.001507147657684982, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.009577441122382879, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.012858473043888808, + "weighted_rkd_loss": 0.0047887205611914395 + }, + { + "epoch": 7.855946398659967, + "flow_matching_loss": 0.014230535924434662, + "grad_norm": 0.13805650174617767, + "learning_rate": 3.162006888127833e-07, + "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": 0.004414458805695176, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0018795684911310673, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.010708486149087548, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.014230535924434662, + "weighted_rkd_loss": 0.005354243074543774 + }, + { + "epoch": 7.872696817420436, + "flow_matching_loss": 0.014107253961265087, + "grad_norm": 0.16493017971515656, + "learning_rate": 2.9625929045422327e-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": 0.0039233009098097685, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0015600109356455505, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.00940661272034049, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.014107253961265087, + "weighted_rkd_loss": 0.004703306360170245 + }, + { + "epoch": 7.889447236180905, + "flow_matching_loss": 0.015400014398619533, + "grad_norm": 0.16884583234786987, + "learning_rate": 2.7696108134781186e-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": 0.0032546573784202337, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0012738020333927125, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.00778311681933701, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.015400014398619533, + "weighted_rkd_loss": 0.003891558409668505 + }, + { + "epoch": 7.906197654941374, + "flow_matching_loss": 0.013906987011432647, + "grad_norm": 0.1894207000732422, + "learning_rate": 2.583069056597376e-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": 0.004085881053470075, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0017150958010461181, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.009886857867240906, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.013906987011432647, + "weighted_rkd_loss": 0.004943428933620453 + }, + { + "epoch": 7.922948073701843, + "flow_matching_loss": 0.011825017677620054, + "grad_norm": 0.16371731460094452, + "learning_rate": 2.4029757938407116e-07, + "loss": 0.0171, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.004440854152198881, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0017637885699514299, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01064549693837762, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.011825017677620054, + "weighted_rkd_loss": 0.00532274846918881 + }, + { + "epoch": 7.939698492462312, + "flow_matching_loss": 0.014789005741477013, + "grad_norm": 0.13796105980873108, + "learning_rate": 2.229338903070871e-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": 0.003778302017599344, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0013750840036664159, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.008931688079610468, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.014789005741477013, + "weighted_rkd_loss": 0.004465844039805234 + }, + { + "epoch": 7.956448911222781, + "flow_matching_loss": 0.013503665756434202, + "grad_norm": 0.13203434646129608, + "learning_rate": 2.0621659797279645e-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": 0.004043603711761534, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0016320106165949254, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.009719217987731099, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.013503665756434202, + "weighted_rkd_loss": 0.004859608993865549 + }, + { + "epoch": 7.97319932998325, + "flow_matching_loss": 0.01366380015388131, + "grad_norm": 0.18163886666297913, + "learning_rate": 1.901464336497233e-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": 0.0043127588927745816, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.001717785291839391, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.010343303112313152, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.01366380015388131, + "weighted_rkd_loss": 0.005171651556156576 + }, + { + "epoch": 7.989949748743719, + "flow_matching_loss": 0.012864657817408443, + "grad_norm": 0.17755556106567383, + "learning_rate": 1.7472410029891527e-07, + "loss": 0.0174, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.0037821149686351418, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0014765637111850082, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.009040793776512146, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.012864657817408443, + "weighted_rkd_loss": 0.004520396888256073 + }, + { + "epoch": 8.006700167504187, + "flow_matching_loss": 0.011925599724054336, + "grad_norm": 0.20086751878261566, + "learning_rate": 1.5995027254319173e-07, + "loss": 0.0161, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.0034393739653751252, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0014759833575226367, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.008354731323197483, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.011925599724054336, + "weighted_rkd_loss": 0.004177365661598742 + }, + { + "epoch": 8.023450586264657, + "flow_matching_loss": 0.013343129586428405, + "grad_norm": 0.12880833446979523, + "learning_rate": 1.4582559663764372e-07, + "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": 0.0033495205221697687, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0017303403757978231, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.00842938139103353, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.013343129586428405, + "weighted_rkd_loss": 0.004214690695516765 + }, + { + "epoch": 8.040201005025125, + "flow_matching_loss": 0.015160613507032395, + "grad_norm": 0.16944606602191925, + "learning_rate": 1.323506904413485e-07, + "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": 0.0036136694019660355, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.001553205365780741, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.00878054415807128, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.015160613507032395, + "weighted_rkd_loss": 0.00439027207903564 + }, + { + "epoch": 8.056951423785595, + "flow_matching_loss": 0.012301782937720418, + "grad_norm": 0.21078728139400482, + "learning_rate": 1.195261433903594e-07, + "loss": 0.0172, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.004058028594590724, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0015868473798036574, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.009702904568985104, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.012301782937720418, + "weighted_rkd_loss": 0.004851452284492552 + }, + { + "epoch": 8.073701842546063, + "flow_matching_loss": 0.012389839068055153, + "grad_norm": 0.141947403550148, + "learning_rate": 1.0735251647190969e-07, + "loss": 0.017, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.0038053673692047597, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.001513481535948813, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.009124216344207525, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.012389839068055153, + "weighted_rkd_loss": 0.004562108172103762 + }, + { + "epoch": 8.090452261306533, + "flow_matching_loss": 0.012021586624905467, + "grad_norm": 0.17158496379852295, + "learning_rate": 9.583034219987408e-08, + "loss": 0.0164, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.003686944651417434, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0014448780042584986, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.008818767266348004, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.012021586624905467, + "weighted_rkd_loss": 0.004409383633174002 + }, + { + "epoch": 8.107202680067001, + "flow_matching_loss": 0.010950662288814783, + "grad_norm": 0.12545715272426605, + "learning_rate": 8.496012459148717e-08, + "loss": 0.016, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.004086814215406775, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0019535994331818076, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01012722784653306, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.010950662288814783, + "weighted_rkd_loss": 0.00506361392326653 + }, + { + "epoch": 8.12395309882747, + "flow_matching_loss": 0.015066776238381863, + "grad_norm": 0.14003939926624298, + "learning_rate": 7.474233914527618e-08, + "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": 0.0033574032597243784, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0013386573467869312, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.008053463837131857, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.015066776238381863, + "weighted_rkd_loss": 0.004026731918565929 + }, + { + "epoch": 8.140703517587939, + "flow_matching_loss": 0.01541524352505803, + "grad_norm": 0.10983029007911682, + "learning_rate": 6.517743282028254e-08, + "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": 0.003956313151866198, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0016391688433941454, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.009551795152947306, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.01541524352505803, + "weighted_rkd_loss": 0.004775897576473653 + }, + { + "epoch": 8.157453936348409, + "flow_matching_loss": 0.0127577087841928, + "grad_norm": 0.145966574549675, + "learning_rate": 5.6265824016492604e-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": 0.004257415048778057, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0016393918893299996, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01015422185882926, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.0127577087841928, + "weighted_rkd_loss": 0.00507711092941463 + }, + { + "epoch": 8.174204355108877, + "flow_matching_loss": 0.0156515721231699, + "grad_norm": 0.12217182666063309, + "learning_rate": 4.800790255654719e-08, + "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": 0.003486564732156694, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0016595104534644634, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.008632639842107891, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.0156515721231699, + "weighted_rkd_loss": 0.004316319921053946 + }, + { + "epoch": 8.190954773869347, + "flow_matching_loss": 0.013703823648393154, + "grad_norm": 0.13512387871742249, + "learning_rate": 4.040402966867862e-08, + "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": 0.004013277473859489, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0014476361509878188, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.009474191255867482, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.013703823648393154, + "weighted_rkd_loss": 0.004737095627933741 + }, + { + "epoch": 8.207705192629815, + "flow_matching_loss": 0.011992666684091092, + "grad_norm": 0.15322472155094147, + "learning_rate": 3.3454537970925014e-08, + "loss": 0.0169, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.003927948325872422, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0020091683545615524, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.009865065244957805, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.011992666684091092, + "weighted_rkd_loss": 0.0049325326224789025 + }, + { + "epoch": 8.224455611390285, + "flow_matching_loss": 0.015503291366621852, + "grad_norm": 0.1615697145462036, + "learning_rate": 2.715973145656192e-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": 0.0033490344532765446, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.001463252492249012, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.008161321375519037, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.015503291366621852, + "weighted_rkd_loss": 0.0040806606877595184 + }, + { + "epoch": 8.241206030150753, + "flow_matching_loss": 0.012654225155711175, + "grad_norm": 0.13093577325344086, + "learning_rate": 2.151988548082129e-08, + "loss": 0.0168, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.00344906086102128, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0013728271587751806, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.008270948939025403, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.012654225155711175, + "weighted_rkd_loss": 0.004135474469512701 + }, + { + "epoch": 8.257956448911223, + "flow_matching_loss": 0.013267476949840784, + "grad_norm": 0.18575097620487213, + "learning_rate": 1.6535246748837772e-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": 0.004015879752114415, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.002008393610594794, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.010040153004229068, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.013267476949840784, + "weighted_rkd_loss": 0.005020076502114534 + }, + { + "epoch": 8.274706867671691, + "flow_matching_loss": 0.013465237710624934, + "grad_norm": 0.1440298706293106, + "learning_rate": 1.2206033304855702e-08, + "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": 0.003453168785199523, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0014610151178203524, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.008367352746427059, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.013465237710624934, + "weighted_rkd_loss": 0.004183676373213529 + }, + { + "epoch": 8.291457286432161, + "flow_matching_loss": 0.012969103502109646, + "grad_norm": 0.12818564474582672, + "learning_rate": 8.532434522695031e-09, + "loss": 0.0179, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.004048713785596192, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0016746725188568235, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.009772100113332271, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.012969103502109646, + "weighted_rkd_loss": 0.004886050056666136 + }, + { + "epoch": 8.308207705192629, + "flow_matching_loss": 0.013686891552060843, + "grad_norm": 0.15679726004600525, + "learning_rate": 5.5146110974679805e-09, + "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": 0.0036764851538464427, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0014912928687408566, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.008844263153150678, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.013686891552060843, + "weighted_rkd_loss": 0.004422131576575339 + }, + { + "epoch": 8.324958123953099, + "flow_matching_loss": 0.01356009249575436, + "grad_norm": 0.1348220556974411, + "learning_rate": 3.1526950385496467e-09, + "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": 0.003303223033435643, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0013120193034410477, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.007918465510010719, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.01356009249575436, + "weighted_rkd_loss": 0.0039592327550053595 + }, + { + "epoch": 8.341708542713567, + "flow_matching_loss": 0.010772908199578523, + "grad_norm": 0.14952634274959564, + "learning_rate": 1.4467896637959666e-09, + "loss": 0.0152, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.003719139681197703, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0014399453764781355, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.008878224762156606, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.010772908199578523, + "weighted_rkd_loss": 0.004439112381078303 + }, + { + "epoch": 8.358458961474037, + "flow_matching_loss": 0.012926962599158287, + "grad_norm": 0.12651483714580536, + "learning_rate": 3.9696959503898733e-10, + "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": 0.003811643342487514, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0014979283267166465, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.009121215017512441, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.012926962599158287, + "weighted_rkd_loss": 0.004560607508756221 + }, + { + "epoch": 8.375209380234505, + "flow_matching_loss": 0.011858177138492465, + "grad_norm": 0.20581090450286865, + "learning_rate": 3.280754806156416e-12, + "loss": 0.017, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.004163221316412091, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0019380579411517828, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.010264500416815281, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.011858177138492465, + "weighted_rkd_loss": 0.005132250208407641 + }, + { + "epoch": 8.375209380234505, + "step": 5000, + "total_flos": 0.0, + "train_loss": 0.03792035340070724, + "train_runtime": 9206.5029, + "train_samples_per_second": 34.758, + "train_steps_per_second": 0.543 + } + ], + "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/seed1/default/train_5shot_seed1_fullactionhead_5k/training_args.bin b/FewShot/39tasks/5shot/FromRKD-Timewarp_FullactionHead/seed1/default/train_5shot_seed1_fullactionhead_5k/training_args.bin new file mode 100644 index 0000000000000000000000000000000000000000..00a6d09a26ddb25b62b00aec37c9a42f80838dce --- /dev/null +++ b/FewShot/39tasks/5shot/FromRKD-Timewarp_FullactionHead/seed1/default/train_5shot_seed1_fullactionhead_5k/training_args.bin @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:4ca330dac47870f5c0b9e52222b90eb78c3ac30f4764dd89362b10daaeda47df +size 6033 diff --git a/FewShot/39tasks/5shot/FromRKD-Timewarp_FullactionHead/seed2/default/train_5shot_seed2_fullactionhead_5k/checkpoint-5000/config.json b/FewShot/39tasks/5shot/FromRKD-Timewarp_FullactionHead/seed2/default/train_5shot_seed2_fullactionhead_5k/checkpoint-5000/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-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-Timewarp_FullactionHead/seed2/default/train_5shot_seed2_fullactionhead_5k/checkpoint-5000/experiment_cfg/metadata.json b/FewShot/39tasks/5shot/FromRKD-Timewarp_FullactionHead/seed2/default/train_5shot_seed2_fullactionhead_5k/checkpoint-5000/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-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": [ + 2.577927505176127, + -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-5000/model-00001-of-00002.safetensors b/FewShot/39tasks/5shot/FromRKD-Timewarp_FullactionHead/seed2/default/train_5shot_seed2_fullactionhead_5k/checkpoint-5000/model-00001-of-00002.safetensors new file mode 100644 index 0000000000000000000000000000000000000000..c90f52dd173a01f62cc39f7ac3fcce55ca525157 --- /dev/null +++ b/FewShot/39tasks/5shot/FromRKD-Timewarp_FullactionHead/seed2/default/train_5shot_seed2_fullactionhead_5k/checkpoint-5000/model-00001-of-00002.safetensors @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:77aaacc7ee85470538018578fecaded4f70641bd8f27d08ebee25a9a605e7a99 +size 4999367032 diff --git a/FewShot/39tasks/5shot/FromRKD-Timewarp_FullactionHead/seed2/default/train_5shot_seed2_fullactionhead_5k/checkpoint-5000/model-00002-of-00002.safetensors b/FewShot/39tasks/5shot/FromRKD-Timewarp_FullactionHead/seed2/default/train_5shot_seed2_fullactionhead_5k/checkpoint-5000/model-00002-of-00002.safetensors new file mode 100644 index 0000000000000000000000000000000000000000..40017848604c545805a73648b52df0f0d03a2059 --- /dev/null +++ b/FewShot/39tasks/5shot/FromRKD-Timewarp_FullactionHead/seed2/default/train_5shot_seed2_fullactionhead_5k/checkpoint-5000/model-00002-of-00002.safetensors @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:e239dee05215eab5d4798f2b6cf01904a026fd99af2c3f053cabc4a4b3f41d74 +size 2664540348 diff --git a/FewShot/39tasks/5shot/FromRKD-Timewarp_FullactionHead/seed2/default/train_5shot_seed2_fullactionhead_5k/checkpoint-5000/model.safetensors.index.json b/FewShot/39tasks/5shot/FromRKD-Timewarp_FullactionHead/seed2/default/train_5shot_seed2_fullactionhead_5k/checkpoint-5000/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-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-Timewarp_FullactionHead/seed2/default/train_5shot_seed2_fullactionhead_5k/checkpoint-5000/optimizer.pt b/FewShot/39tasks/5shot/FromRKD-Timewarp_FullactionHead/seed2/default/train_5shot_seed2_fullactionhead_5k/checkpoint-5000/optimizer.pt new file mode 100644 index 0000000000000000000000000000000000000000..ca15173cb8edf93d4c8a9b095e37518bdb6e9541 --- /dev/null +++ b/FewShot/39tasks/5shot/FromRKD-Timewarp_FullactionHead/seed2/default/train_5shot_seed2_fullactionhead_5k/checkpoint-5000/optimizer.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:a19dbc17222c8452fc2f1b373a4cba22d0aea3a7022f8f94dcd55ff210982b33 +size 8563326263 diff --git a/FewShot/39tasks/5shot/FromRKD-Timewarp_FullactionHead/seed2/default/train_5shot_seed2_fullactionhead_5k/checkpoint-5000/rng_state.pth b/FewShot/39tasks/5shot/FromRKD-Timewarp_FullactionHead/seed2/default/train_5shot_seed2_fullactionhead_5k/checkpoint-5000/rng_state.pth new file mode 100644 index 0000000000000000000000000000000000000000..81328f76ddb63b8e3080ac985f95e39e14013ddc --- /dev/null +++ b/FewShot/39tasks/5shot/FromRKD-Timewarp_FullactionHead/seed2/default/train_5shot_seed2_fullactionhead_5k/checkpoint-5000/rng_state.pth @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:52d8b7794775502c36c9f18edf928415953e0f0e15962e98a47a94f03bf16dba +size 14645 diff --git a/FewShot/39tasks/5shot/FromRKD-Timewarp_FullactionHead/seed2/default/train_5shot_seed2_fullactionhead_5k/checkpoint-5000/scheduler.pt b/FewShot/39tasks/5shot/FromRKD-Timewarp_FullactionHead/seed2/default/train_5shot_seed2_fullactionhead_5k/checkpoint-5000/scheduler.pt new file mode 100644 index 0000000000000000000000000000000000000000..8cb18c42a19c524a920dd1ec232118cfa1c2aa5a --- /dev/null +++ b/FewShot/39tasks/5shot/FromRKD-Timewarp_FullactionHead/seed2/default/train_5shot_seed2_fullactionhead_5k/checkpoint-5000/scheduler.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:39f93e93eb5653777630c590a20e98e9373f48f53da024592383b624fc2dcbde +size 1465 diff --git a/FewShot/39tasks/5shot/FromRKD-Timewarp_FullactionHead/seed2/default/train_5shot_seed2_fullactionhead_5k/checkpoint-5000/trainer_state.json b/FewShot/39tasks/5shot/FromRKD-Timewarp_FullactionHead/seed2/default/train_5shot_seed2_fullactionhead_5k/checkpoint-5000/trainer_state.json new file mode 100644 index 0000000000000000000000000000000000000000..83e4b4a8bd45a14b3b48956114c7ec10cb68c9bf --- /dev/null +++ b/FewShot/39tasks/5shot/FromRKD-Timewarp_FullactionHead/seed2/default/train_5shot_seed2_fullactionhead_5k/checkpoint-5000/trainer_state.json @@ -0,0 +1,23034 @@ +{ + "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": 0.1910901501774788, + "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 + }, + { + "epoch": 5.145299145299146, + "flow_matching_loss": 0.015014776308089495, + "grad_norm": 0.18950475752353668, + "learning_rate": 1.1231228239056467e-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": 0.005329322838224471, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0022741535911336543, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.012932799477130174, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.015014776308089495, + "weighted_rkd_loss": 0.006466399738565087 + }, + { + "epoch": 5.162393162393163, + "flow_matching_loss": 0.015901549346745016, + "grad_norm": 0.14783582091331482, + "learning_rate": 1.113528559434845e-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": 0.005319552915170789, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.002164692198857665, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.012803797982633114, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.015901549346745016, + "weighted_rkd_loss": 0.006401898991316557 + }, + { + "epoch": 5.17948717948718, + "flow_matching_loss": 0.016365189291536807, + "grad_norm": 0.15554279088974, + "learning_rate": 1.1039512004768476e-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": 0.004656393500044942, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.002018486650194973, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.011331273661926388, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.016365189291536807, + "weighted_rkd_loss": 0.005665636830963194 + }, + { + "epoch": 5.196581196581197, + "flow_matching_loss": 0.01562155494466424, + "grad_norm": 0.19159680604934692, + "learning_rate": 1.0943911659763612e-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": 0.005035868706181645, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.002107031480409205, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.012178768776357174, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.01562155494466424, + "weighted_rkd_loss": 0.006089384388178587 + }, + { + "epoch": 5.213675213675214, + "flow_matching_loss": 0.015252807457000018, + "grad_norm": 0.1729981005191803, + "learning_rate": 1.084848874120266e-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": 0.0056610606145113705, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0025155096664093435, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.013837630953639745, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.015252807457000018, + "weighted_rkd_loss": 0.006918815476819873 + }, + { + "epoch": 5.230769230769231, + "flow_matching_loss": 0.017009658738970757, + "grad_norm": 0.20856738090515137, + "learning_rate": 1.0753247423193196e-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": 0.005203615548089147, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0021950440714135766, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.012602275051176549, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.017009658738970757, + "weighted_rkd_loss": 0.006301137525588274 + }, + { + "epoch": 5.247863247863248, + "flow_matching_loss": 0.014320117607712745, + "grad_norm": 0.200138121843338, + "learning_rate": 1.0658191871899012e-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": 0.004975240374915302, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0019149117753840983, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.011865392327308655, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.014320117607712745, + "weighted_rkd_loss": 0.005932696163654327 + }, + { + "epoch": 5.264957264957265, + "flow_matching_loss": 0.01623127833008766, + "grad_norm": 0.1671903431415558, + "learning_rate": 1.0563326245357852e-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": 0.005758790764957666, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.002532364521175623, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.014049946237355471, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.01623127833008766, + "weighted_rkd_loss": 0.0070249731186777355 + }, + { + "epoch": 5.282051282051282, + "flow_matching_loss": 0.013452016282826662, + "grad_norm": 0.1713135540485382, + "learning_rate": 1.0468654693299557e-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": 0.0055105118080973625, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0022663405281491576, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01328736413270235, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.013452016282826662, + "weighted_rkd_loss": 0.006643682066351175 + }, + { + "epoch": 5.299145299145299, + "flow_matching_loss": 0.015398501604795455, + "grad_norm": 0.15591202676296234, + "learning_rate": 1.0374181356964501e-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": 0.005339693161658943, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.002155479183420539, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.012834865320473909, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.015398501604795455, + "weighted_rkd_loss": 0.006417432660236954 + }, + { + "epoch": 5.316239316239316, + "flow_matching_loss": 0.013474423624575137, + "grad_norm": 0.186759814620018, + "learning_rate": 1.0279910368922465e-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": 0.005282991589047015, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.00228860656498, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.012854589708149433, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.013474423624575137, + "weighted_rkd_loss": 0.006427294854074716 + }, + { + "epoch": 5.333333333333333, + "flow_matching_loss": 0.014557658508419991, + "grad_norm": 0.19690386950969696, + "learning_rate": 1.018584585289186e-05, + "loss": 0.021, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.005379770835861564, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0021475294954143463, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01290707103908062, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.014557658508419991, + "weighted_rkd_loss": 0.00645353551954031 + }, + { + "epoch": 5.35042735042735, + "flow_matching_loss": 0.016512978728860615, + "grad_norm": 0.24174918234348297, + "learning_rate": 1.0091991923559348e-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": 0.005569035932421684, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0023933482705615463, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.013531420193612576, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.016512978728860615, + "weighted_rkd_loss": 0.006765710096806288 + }, + { + "epoch": 5.367521367521368, + "flow_matching_loss": 0.015567683707922697, + "grad_norm": 0.17910248041152954, + "learning_rate": 9.99835268639984e-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": 0.005675594275817275, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0023775284411385656, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.013728717155754567, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.015567683707922697, + "weighted_rkd_loss": 0.0068643585778772834 + }, + { + "epoch": 5.384615384615385, + "flow_matching_loss": 0.012863228842616082, + "grad_norm": 0.17725598812103271, + "learning_rate": 9.904932237496913e-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": 0.004814130300655961, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.002020637388341129, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.011648897919803858, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.012863228842616082, + "weighted_rkd_loss": 0.005824448959901929 + }, + { + "epoch": 5.401709401709402, + "flow_matching_loss": 0.012962156720459462, + "grad_norm": 0.21131287515163422, + "learning_rate": 9.811734663363659e-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": 0.005019215657375753, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0023267841315828264, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.012365215551108122, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.012962156720459462, + "weighted_rkd_loss": 0.006182607775554061 + }, + { + "epoch": 5.418803418803419, + "flow_matching_loss": 0.016335158282890914, + "grad_norm": 0.16600927710533142, + "learning_rate": 9.718764040763875e-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": 0.005628416361287236, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0027493985020555557, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.014006231166422366, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.016335158282890914, + "weighted_rkd_loss": 0.007003115583211183 + }, + { + "epoch": 5.435897435897436, + "flow_matching_loss": 0.012864034436643124, + "grad_norm": 0.2108919471502304, + "learning_rate": 9.626024436533777e-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": 0.0059022421948611735, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0023489799350500107, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.014153464417904615, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.012864034436643124, + "weighted_rkd_loss": 0.007076732208952307 + }, + { + "epoch": 5.452991452991453, + "flow_matching_loss": 0.014886892028152942, + "grad_norm": 0.21139591932296753, + "learning_rate": 9.533519907404095e-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": 0.005289020924828946, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.002200934721622616, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.012778976373374462, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.014886892028152942, + "weighted_rkd_loss": 0.006389488186687231 + }, + { + "epoch": 5.47008547008547, + "flow_matching_loss": 0.016243355721235274, + "grad_norm": 0.13592956960201263, + "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": 0.005620071780867875, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.002551423793192953, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.013791567226871849, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.016243355721235274, + "weighted_rkd_loss": 0.006895783613435924 + }, + { + "epoch": 5.487179487179487, + "flow_matching_loss": 0.015402343589812517, + "grad_norm": 0.14428843557834625, + "learning_rate": 9.349232249777124e-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": 0.005076029896736145, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.002067275461740792, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.012219335231930018, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.015402343589812517, + "weighted_rkd_loss": 0.006109667615965009 + }, + { + "epoch": 5.504273504273504, + "flow_matching_loss": 0.01771446131169796, + "grad_norm": 0.16076010465621948, + "learning_rate": 9.257457182618985e-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": 0.005623388360254467, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0022086908458732067, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.013455467531457543, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.01771446131169796, + "weighted_rkd_loss": 0.006727733765728772 + }, + { + "epoch": 5.521367521367521, + "flow_matching_loss": 0.014067314937710762, + "grad_norm": 0.17357373237609863, + "learning_rate": 9.165933312886944e-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": 0.005197142693214118, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0022083621006458998, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.012602647487074136, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.014067314937710762, + "weighted_rkd_loss": 0.006301323743537068 + }, + { + "epoch": 5.538461538461538, + "flow_matching_loss": 0.014604238979518413, + "grad_norm": 0.19311626255512238, + "learning_rate": 9.07466464413155e-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": 0.004733706871047616, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0022366787889041007, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.011704092426225544, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.014604238979518413, + "weighted_rkd_loss": 0.005852046213112772 + }, + { + "epoch": 5.555555555555555, + "flow_matching_loss": 0.015962645784020423, + "grad_norm": 0.1606195569038391, + "learning_rate": 8.983655168740048e-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": 0.004856789810582995, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0018019488430581986, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.011515528429299593, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.015962645784020423, + "weighted_rkd_loss": 0.0057577642146497965 + }, + { + "epoch": 5.572649572649572, + "flow_matching_loss": 0.014079044666141271, + "grad_norm": 0.18407145142555237, + "learning_rate": 8.892908867761718e-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": 0.005349058820866048, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0023212913540191947, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.013019408844411374, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.014079044666141271, + "weighted_rkd_loss": 0.006509704422205687 + }, + { + "epoch": 5.589743589743589, + "flow_matching_loss": 0.014307726826518773, + "grad_norm": 0.1682978719472885, + "learning_rate": 8.802429710733745e-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": 0.005156373279169202, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0019659399753436447, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01227868665009737, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.014307726826518773, + "weighted_rkd_loss": 0.006139343325048685 + }, + { + "epoch": 5.6068376068376065, + "flow_matching_loss": 0.014214032515883445, + "grad_norm": 0.1408603936433792, + "learning_rate": 8.71222165550757e-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": 0.004618929629214108, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0018305400968529283, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.011068399343639613, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.014214032515883445, + "weighted_rkd_loss": 0.0055341996718198064 + }, + { + "epoch": 5.6239316239316235, + "flow_matching_loss": 0.01317118015140295, + "grad_norm": 0.13524463772773743, + "learning_rate": 8.622288648075748e-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": 0.0049630893394351, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0020186342881061137, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.011944812955334783, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.01317118015140295, + "weighted_rkd_loss": 0.005972406477667391 + }, + { + "epoch": 5.641025641025641, + "flow_matching_loss": 0.016691511496901512, + "grad_norm": 0.13098731637001038, + "learning_rate": 8.532634622399391e-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": 0.004113830532878638, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0016622172726783902, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.009889878332614899, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.016691511496901512, + "weighted_rkd_loss": 0.004944939166307449 + }, + { + "epoch": 5.6581196581196584, + "flow_matching_loss": 0.013253772258758545, + "grad_norm": 0.1594141125679016, + "learning_rate": 8.44326350023601e-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": 0.00496185205411166, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0018524963175877928, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.011776200449094176, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.013253772258758545, + "weighted_rkd_loss": 0.005888100224547088 + }, + { + "epoch": 5.6752136752136755, + "flow_matching_loss": 0.01386769926175475, + "grad_norm": 0.18667681515216827, + "learning_rate": 8.354179190968002e-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": 0.0042942218715325, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0017211545608006417, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.010309598175808787, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.01386769926175475, + "weighted_rkd_loss": 0.005154799087904394 + }, + { + "epoch": 5.6923076923076925, + "flow_matching_loss": 0.015427383687347173, + "grad_norm": 0.19144976139068604, + "learning_rate": 8.265385591431661e-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": 0.004844757309183478, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.001974588306620717, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.011664103018119931, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.015427383687347173, + "weighted_rkd_loss": 0.005832051509059966 + }, + { + "epoch": 5.7094017094017095, + "flow_matching_loss": 0.014562000147998333, + "grad_norm": 0.181470587849617, + "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": 0.005005742516368627, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.001948601473122835, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.011960086273029446, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.014562000147998333, + "weighted_rkd_loss": 0.005980043136514723 + }, + { + "epoch": 5.726495726495727, + "flow_matching_loss": 0.01286997003480792, + "grad_norm": 0.2129090130329132, + "learning_rate": 8.088686045146247e-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": 0.00470130869653076, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0017387204454280436, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.011141337687149644, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.01286997003480792, + "weighted_rkd_loss": 0.005570668843574822 + }, + { + "epoch": 5.743589743589744, + "flow_matching_loss": 0.013654280174523591, + "grad_norm": 0.17266348004341125, + "learning_rate": 8.00078782780775e-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": 0.005098359915427864, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.002111944928765297, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.012308664619922638, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.013654280174523591, + "weighted_rkd_loss": 0.006154332309961319 + }, + { + "epoch": 5.760683760683761, + "flow_matching_loss": 0.016567762661725282, + "grad_norm": 0.22100067138671875, + "learning_rate": 7.913195778683979e-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": 0.00517838706728071, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.001889878825750202, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.012246652832254767, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.016567762661725282, + "weighted_rkd_loss": 0.0061233264161273835 + }, + { + "epoch": 5.777777777777778, + "flow_matching_loss": 0.014265890140086413, + "grad_norm": 0.1820782721042633, + "learning_rate": 7.825913729334908e-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": 0.004978736140765249, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.001999659137800336, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.011957131465896964, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.014265890140086413, + "weighted_rkd_loss": 0.005978565732948482 + }, + { + "epoch": 5.794871794871795, + "flow_matching_loss": 0.016519154980778696, + "grad_norm": 0.17449019849300385, + "learning_rate": 7.73894549776013e-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": 0.004733387008309364, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0020300828851759433, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.011496856901794672, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.016519154980778696, + "weighted_rkd_loss": 0.005748428450897336 + }, + { + "epoch": 5.811965811965812, + "flow_matching_loss": 0.01309665311127901, + "grad_norm": 0.1982821375131607, + "learning_rate": 7.65229488823184e-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": 0.00616358332335949, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.002463771589100361, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.014790938142687083, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.01309665311127901, + "weighted_rkd_loss": 0.007395469071343541 + }, + { + "epoch": 5.829059829059829, + "flow_matching_loss": 0.01534111937507987, + "grad_norm": 0.16873861849308014, + "learning_rate": 7.565965691128396e-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": 0.005616517853923142, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0022936845547519625, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01352672027423978, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.01534111937507987, + "weighted_rkd_loss": 0.00676336013711989 + }, + { + "epoch": 5.846153846153846, + "flow_matching_loss": 0.014631201140582561, + "grad_norm": 0.19740276038646698, + "learning_rate": 7.47996168276856e-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": 0.005165145918726921, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0019399866694584488, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.012270278483629226, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.014631201140582561, + "weighted_rkd_loss": 0.006135139241814613 + }, + { + "epoch": 5.863247863247864, + "flow_matching_loss": 0.014703481085598468, + "grad_norm": 0.1830338090658188, + "learning_rate": 7.394286625246258e-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": 0.0049274742603302, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0022761092172004283, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.012131057772785425, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.014703481085598468, + "weighted_rkd_loss": 0.006065528886392712 + }, + { + "epoch": 5.880341880341881, + "flow_matching_loss": 0.013996908906847239, + "grad_norm": 0.12798833847045898, + "learning_rate": 7.308944266266063e-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": 0.00495609906502068, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0020264101680368186, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01193860825151205, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.013996908906847239, + "weighted_rkd_loss": 0.005969304125756025 + }, + { + "epoch": 5.897435897435898, + "flow_matching_loss": 0.015623558219522238, + "grad_norm": 0.17718824744224548, + "learning_rate": 7.223938338979233e-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": 0.004302141885273158, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0017569319461472332, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.010361215798184276, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.015623558219522238, + "weighted_rkd_loss": 0.005180607899092138 + }, + { + "epoch": 5.914529914529915, + "flow_matching_loss": 0.015388756524771452, + "grad_norm": 0.1666608452796936, + "learning_rate": 7.139272561820397e-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": 0.004684569523669779, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0017473775544203818, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01111651649698615, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.015388756524771452, + "weighted_rkd_loss": 0.005558258248493075 + }, + { + "epoch": 5.931623931623932, + "flow_matching_loss": 0.013417072873562575, + "grad_norm": 0.13867004215717316, + "learning_rate": 7.054950638344936e-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": 0.005452374718151987, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0022980000008828937, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.013202749611809849, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.013417072873562575, + "weighted_rkd_loss": 0.0066013748059049245 + }, + { + "epoch": 5.948717948717949, + "flow_matching_loss": 0.012726581562310458, + "grad_norm": 0.14203988015651703, + "learning_rate": 6.970976257066941e-06, + "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": 0.003928991034626961, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0017228475771844387, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.00958082964643836, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.012726581562310458, + "weighted_rkd_loss": 0.00479041482321918 + }, + { + "epoch": 5.965811965811966, + "flow_matching_loss": 0.014724280033260585, + "grad_norm": 0.17691625654697418, + "learning_rate": 6.887353091297896e-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": 0.005078516318462789, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0020477183861657977, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01220475104637444, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.014724280033260585, + "weighted_rkd_loss": 0.00610237552318722 + }, + { + "epoch": 5.982905982905983, + "flow_matching_loss": 0.013537014741450549, + "grad_norm": 0.15324710309505463, + "learning_rate": 6.804084798985965e-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": 0.00474430185277015, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0019807930337265136, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01146939666941762, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.013537014741450549, + "weighted_rkd_loss": 0.00573469833470881 + }, + { + "epoch": 6.0, + "flow_matching_loss": 0.015401409938931465, + "grad_norm": 0.17201311886310577, + "learning_rate": 6.721175022556009e-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": 0.004189198603853583, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0016246947110630573, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01000309200026095, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.015401409938931465, + "weighted_rkd_loss": 0.005001546000130475 + }, + { + "epoch": 6.017094017094017, + "flow_matching_loss": 0.01367185264825821, + "grad_norm": 0.1638801097869873, + "learning_rate": 6.638627388750245e-06, + "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": 0.004297086899168789, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.001833447511307895, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01042762128636241, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.01367185264825821, + "weighted_rkd_loss": 0.005213810643181205 + }, + { + "epoch": 6.034188034188034, + "flow_matching_loss": 0.015124630182981491, + "grad_norm": 0.15200188755989075, + "learning_rate": 6.556445508469583e-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": 0.0053081318968906995, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0021283041802234947, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.012744567822664976, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.015124630182981491, + "weighted_rkd_loss": 0.006372283911332488 + }, + { + "epoch": 6.051282051282051, + "flow_matching_loss": 0.012258633133023977, + "grad_norm": 0.170354962348938, + "learning_rate": 6.474632976615708e-06, + "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": 0.0043724456802010534, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0018084332928992807, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.010553324595093727, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.012258633133023977, + "weighted_rkd_loss": 0.0052766622975468636 + }, + { + "epoch": 6.068376068376068, + "flow_matching_loss": 0.014518317673355341, + "grad_norm": 0.18661901354789734, + "learning_rate": 6.393193371933791e-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": 0.00458554292563349, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0020006953971460463, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.011171781178563833, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.014518317673355341, + "weighted_rkd_loss": 0.005585890589281916 + }, + { + "epoch": 6.085470085470085, + "flow_matching_loss": 0.012813680432736873, + "grad_norm": 0.17715997993946075, + "learning_rate": 6.312130256855977e-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": 0.004458332341164351, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.001910470041912049, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01082713482901454, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.012813680432736873, + "weighted_rkd_loss": 0.00541356741450727 + }, + { + "epoch": 6.102564102564102, + "flow_matching_loss": 0.01346785482019186, + "grad_norm": 0.16243726015090942, + "learning_rate": 6.23144717734552e-06, + "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": 0.004544325242750347, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0018196132208686323, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.010908263735473156, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.01346785482019186, + "weighted_rkd_loss": 0.005454131867736578 + }, + { + "epoch": 6.119658119658119, + "flow_matching_loss": 0.014472407475113868, + "grad_norm": 0.17122097313404083, + "learning_rate": 6.151147662741695e-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": 0.004576569981873035, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.001975224877241999, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.011128364969044924, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.014472407475113868, + "weighted_rkd_loss": 0.005564182484522462 + }, + { + "epoch": 6.136752136752137, + "flow_matching_loss": 0.014526502415537833, + "grad_norm": 0.1407535821199417, + "learning_rate": 6.071235225605415e-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": 0.004797903215512633, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0018253034097142518, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.011421109968796372, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.014526502415537833, + "weighted_rkd_loss": 0.005710554984398186 + }, + { + "epoch": 6.153846153846154, + "flow_matching_loss": 0.013544728001579642, + "grad_norm": 0.13234646618366241, + "learning_rate": 5.991713361565546e-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": 0.0049188114702701565, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0021697683492675423, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.012007391452789307, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.013544728001579642, + "weighted_rkd_loss": 0.0060036957263946535 + }, + { + "epoch": 6.170940170940171, + "flow_matching_loss": 0.013128768280148506, + "grad_norm": 0.17086155712604523, + "learning_rate": 5.912585549166051e-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": 0.004967864602804184, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0018786639207974075, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.011814393242821098, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.013128768280148506, + "weighted_rkd_loss": 0.005907196621410549 + }, + { + "epoch": 6.188034188034188, + "flow_matching_loss": 0.013461261987686157, + "grad_norm": 0.17070406675338745, + "learning_rate": 5.833855249713777e-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": 0.004719796217978, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.001841046300251037, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.011280638724565506, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.013461261987686157, + "weighted_rkd_loss": 0.005640319362282753 + }, + { + "epoch": 6.205128205128205, + "flow_matching_loss": 0.014641128573566675, + "grad_norm": 0.19366569817066193, + "learning_rate": 5.7555259071270645e-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": 0.004503894480876625, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0017051879549399018, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.010712976986542343, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.014641128573566675, + "weighted_rkd_loss": 0.005356488493271172 + }, + { + "epoch": 6.222222222222222, + "flow_matching_loss": 0.012268691649660468, + "grad_norm": 0.17229615151882172, + "learning_rate": 5.677600947785117e-06, + "loss": 0.0179, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.004729417897760868, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0018767193774692713, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.011335555091500282, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.012268691649660468, + "weighted_rkd_loss": 0.005667777545750141 + }, + { + "epoch": 6.239316239316239, + "flow_matching_loss": 0.013378062937408685, + "grad_norm": 0.18267109990119934, + "learning_rate": 5.600083780378099e-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": 0.004501617699861527, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.001768733304925263, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.010771968681365252, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.013378062937408685, + "weighted_rkd_loss": 0.005385984340682626 + }, + { + "epoch": 6.256410256410256, + "flow_matching_loss": 0.01337980325333774, + "grad_norm": 0.2579333782196045, + "learning_rate": 5.522977795758038e-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": 0.004270854382775724, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.001658515038434416, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.010200223792344332, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.01337980325333774, + "weighted_rkd_loss": 0.005100111896172166 + }, + { + "epoch": 6.273504273504273, + "flow_matching_loss": 0.014495591539889574, + "grad_norm": 0.19292736053466797, + "learning_rate": 5.4462863667904805e-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": 0.005087881651706994, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0017388155218213796, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.011914578825235366, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.014495591539889574, + "weighted_rkd_loss": 0.005957289412617683 + }, + { + "epoch": 6.2905982905982905, + "flow_matching_loss": 0.014014208316802978, + "grad_norm": 0.2239702343940735, + "learning_rate": 5.370012848206983e-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": 0.004500076500698924, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0018213755334727465, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.010821528499945998, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.014014208316802978, + "weighted_rkd_loss": 0.005410764249972999 + }, + { + "epoch": 6.3076923076923075, + "flow_matching_loss": 0.0140917822252959, + "grad_norm": 0.17948734760284424, + "learning_rate": 5.29416057645834e-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": 0.0042427378240972756, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0016766243032179772, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.010162099963054061, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.0140917822252959, + "weighted_rkd_loss": 0.0050810499815270305 + }, + { + "epoch": 6.3247863247863245, + "flow_matching_loss": 0.014403108693659305, + "grad_norm": 0.15205460786819458, + "learning_rate": 5.218732869568635e-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": 0.0042243374977260825, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0018360894406214356, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.010284764459356666, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.014403108693659305, + "weighted_rkd_loss": 0.005142382229678333 + }, + { + "epoch": 6.3418803418803416, + "flow_matching_loss": 0.012675300985574723, + "grad_norm": 0.1909080594778061, + "learning_rate": 5.14373302699013e-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": 0.005056148441508412, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.00219099071691744, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.012303287675604225, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.012675300985574723, + "weighted_rkd_loss": 0.0061516438378021124 + }, + { + "epoch": 6.358974358974359, + "flow_matching_loss": 0.014092040248215199, + "grad_norm": 0.17801937460899353, + "learning_rate": 5.069164329458894e-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": 0.004562873602844774, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0017242739209905266, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.010850021056830883, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.014092040248215199, + "weighted_rkd_loss": 0.005425010528415442 + }, + { + "epoch": 6.3760683760683765, + "flow_matching_loss": 0.014610704127699136, + "grad_norm": 0.16385993361473083, + "learning_rate": 4.995030038851342e-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": 0.004508930863812566, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0017317418241873384, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.010749603435397149, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.014610704127699136, + "weighted_rkd_loss": 0.005374801717698574 + }, + { + "epoch": 6.3931623931623935, + "flow_matching_loss": 0.011634174268692732, + "grad_norm": 0.15849269926548004, + "learning_rate": 4.9213333980415e-06, + "loss": 0.0166, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.004152471316047013, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.001584779960103333, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.009889722522348165, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.011634174268692732, + "weighted_rkd_loss": 0.004944861261174082 + }, + { + "epoch": 6.410256410256411, + "flow_matching_loss": 0.013286026380956172, + "grad_norm": 0.1140068992972374, + "learning_rate": 4.848077630759175e-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": 0.00479142447002232, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0021043072978500278, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.011687156278640032, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.013286026380956172, + "weighted_rkd_loss": 0.005843578139320016 + }, + { + "epoch": 6.427350427350428, + "flow_matching_loss": 0.014323109947144985, + "grad_norm": 0.17702610790729523, + "learning_rate": 4.775265941448951e-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": 0.005099014192819595, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0019039258710108697, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.012101954314857721, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.014323109947144985, + "weighted_rkd_loss": 0.006050977157428861 + }, + { + "epoch": 6.444444444444445, + "flow_matching_loss": 0.013860237039625644, + "grad_norm": 0.1613358110189438, + "learning_rate": 4.702901515129989e-06, + "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": 0.004091036505997181, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.001842647383455187, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.010024720383808017, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.013860237039625644, + "weighted_rkd_loss": 0.005012360191904009 + }, + { + "epoch": 6.461538461538462, + "flow_matching_loss": 0.013166516833007336, + "grad_norm": 0.12258017063140869, + "learning_rate": 4.630987517256734e-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": 0.005356473685242236, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.00200028627878055, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.012713233754038811, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.013166516833007336, + "weighted_rkd_loss": 0.006356616877019406 + }, + { + "epoch": 6.478632478632479, + "flow_matching_loss": 0.012790456414222717, + "grad_norm": 0.1847393661737442, + "learning_rate": 4.559527093580412e-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": 0.004619638319127262, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0019545887014828624, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.011193865444511175, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.012790456414222717, + "weighted_rkd_loss": 0.005596932722255588 + }, + { + "epoch": 6.495726495726496, + "flow_matching_loss": 0.012947164382785559, + "grad_norm": 0.16731011867523193, + "learning_rate": 4.488523370011456e-06, + "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": 0.004310220014303923, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0016456021461635829, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.010266042361035944, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.012947164382785559, + "weighted_rkd_loss": 0.005133021180517972 + }, + { + "epoch": 6.512820512820513, + "flow_matching_loss": 0.012447025114670395, + "grad_norm": 0.13510732352733612, + "learning_rate": 4.417979452482766e-06, + "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": 0.004228096292354167, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0016668969125021248, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.010123089654371142, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.012447025114670395, + "weighted_rkd_loss": 0.005061544827185571 + }, + { + "epoch": 6.52991452991453, + "flow_matching_loss": 0.01330544799566269, + "grad_norm": 0.14951033890247345, + "learning_rate": 4.347898426813811e-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": 0.00396587192080915, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0014627483498770744, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.009394492162391544, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.01330544799566269, + "weighted_rkd_loss": 0.004697246081195772 + }, + { + "epoch": 6.547008547008547, + "flow_matching_loss": 0.01360499719157815, + "grad_norm": 0.16325245797634125, + "learning_rate": 4.278283358575694e-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": 0.004816550016403198, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0018093404709361494, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.011442440561950207, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.01360499719157815, + "weighted_rkd_loss": 0.005721220280975103 + }, + { + "epoch": 6.564102564102564, + "flow_matching_loss": 0.01414185781031847, + "grad_norm": 0.2390795648097992, + "learning_rate": 4.2091372929570145e-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": 0.004850478307344019, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.002294772188179195, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.011995728733018041, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.01414185781031847, + "weighted_rkd_loss": 0.0059978643665090205 + }, + { + "epoch": 6.581196581196581, + "flow_matching_loss": 0.016646978538483383, + "grad_norm": 0.1479363888502121, + "learning_rate": 4.1404632546306934e-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": 0.0038240491645410656, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0016426807967945933, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.009290779009461403, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.016646978538483383, + "weighted_rkd_loss": 0.0046453895047307014 + }, + { + "epoch": 6.598290598290598, + "flow_matching_loss": 0.013259697053581477, + "grad_norm": 0.1548348218202591, + "learning_rate": 4.0722642476216286e-06, + "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": 0.004015104891732335, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0015654072863981127, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.00959561700001359, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.013259697053581477, + "weighted_rkd_loss": 0.004797808500006795 + }, + { + "epoch": 6.615384615384615, + "flow_matching_loss": 0.013232738617807627, + "grad_norm": 0.17228160798549652, + "learning_rate": 4.00454325517533e-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": 0.0046642918838188056, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0017990636872127653, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.011127647291868926, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.013232738617807627, + "weighted_rkd_loss": 0.005563823645934463 + }, + { + "epoch": 6.632478632478632, + "flow_matching_loss": 0.013004853390157223, + "grad_norm": 0.17855753004550934, + "learning_rate": 3.937303239627393e-06, + "loss": 0.017, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.003314048214815557, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0014266120444517583, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.008054708503186703, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.013004853390157223, + "weighted_rkd_loss": 0.0040273542515933515 + }, + { + "epoch": 6.64957264957265, + "flow_matching_loss": 0.014002307411283254, + "grad_norm": 0.24588173627853394, + "learning_rate": 3.870547142273918e-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": 0.00451577752828598, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0020319839182775467, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.011063539003953338, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.014002307411283254, + "weighted_rkd_loss": 0.005531769501976669 + }, + { + "epoch": 6.666666666666667, + "flow_matching_loss": 0.013700271677225827, + "grad_norm": 0.1725107878446579, + "learning_rate": 3.8042778832428736e-06, + "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": 0.0036921764723956583, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0014130157709587366, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.008797368640080095, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.013700271677225827, + "weighted_rkd_loss": 0.004398684320040047 + }, + { + "epoch": 6.683760683760684, + "flow_matching_loss": 0.01752990819513798, + "grad_norm": 0.1484254151582718, + "learning_rate": 3.7384983613663195e-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": 0.004090518108569086, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0015098612115252764, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.00969089725986123, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.01752990819513798, + "weighted_rkd_loss": 0.004845448629930615 + }, + { + "epoch": 6.700854700854701, + "flow_matching_loss": 0.015219774376600981, + "grad_norm": 0.21970634162425995, + "learning_rate": 3.6732114540536503e-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": 0.004155363631434738, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0016006131307221949, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.009911340568214655, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.015219774376600981, + "weighted_rkd_loss": 0.004955670284107328 + }, + { + "epoch": 6.717948717948718, + "flow_matching_loss": 0.014597513433545827, + "grad_norm": 0.1565510630607605, + "learning_rate": 3.608420017165683e-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": 0.004712637630291283, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0017746063880622388, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.011199881695210934, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.014597513433545827, + "weighted_rkd_loss": 0.005599940847605467 + }, + { + "epoch": 6.735042735042735, + "flow_matching_loss": 0.015428585838526488, + "grad_norm": 0.21946097910404205, + "learning_rate": 3.5441268848897664e-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": 0.004173310683108866, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0015550711716059596, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.009901692438870669, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.015428585838526488, + "weighted_rkd_loss": 0.0049508462194353346 + }, + { + "epoch": 6.752136752136752, + "flow_matching_loss": 0.012527666334062815, + "grad_norm": 0.13608911633491516, + "learning_rate": 3.480334869615792e-06, + "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": 0.004187409998849034, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.001549099455587566, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.009923919336870313, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.012527666334062815, + "weighted_rkd_loss": 0.004961959668435156 + }, + { + "epoch": 6.769230769230769, + "flow_matching_loss": 0.01083629266358912, + "grad_norm": 0.18144173920154572, + "learning_rate": 3.417046761813155e-06, + "loss": 0.0163, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.004622485325671733, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0017338956182356923, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.010978866182267667, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.01083629266358912, + "weighted_rkd_loss": 0.005489433091133833 + }, + { + "epoch": 6.786324786324786, + "flow_matching_loss": 0.014319020509719848, + "grad_norm": 0.21980780363082886, + "learning_rate": 3.3542653299087263e-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": 0.004227689397521317, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0015997855283785612, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.010055164387449623, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.014319020509719848, + "weighted_rkd_loss": 0.005027582193724811 + }, + { + "epoch": 6.803418803418803, + "flow_matching_loss": 0.010222503775730729, + "grad_norm": 0.1556553691625595, + "learning_rate": 3.2919933201657117e-06, + "loss": 0.0148, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.0037521095015108584, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0016987596172839403, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.009202978760004043, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.010222503775730729, + "weighted_rkd_loss": 0.0046014893800020214 + }, + { + "epoch": 6.82051282051282, + "flow_matching_loss": 0.014451631903648376, + "grad_norm": 0.19691146910190582, + "learning_rate": 3.2302334565635584e-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": 0.004138553678058088, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0016731622163206339, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.009950269386172295, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.014451631903648376, + "weighted_rkd_loss": 0.0049751346930861475 + }, + { + "epoch": 6.837606837606837, + "flow_matching_loss": 0.011851238133385778, + "grad_norm": 0.20504842698574066, + "learning_rate": 3.1689884406787674e-06, + "loss": 0.0163, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.0037106047151610256, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.001522247534012422, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.008943456877022981, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.011851238133385778, + "weighted_rkd_loss": 0.0044717284385114905 + }, + { + "epoch": 6.854700854700854, + "flow_matching_loss": 0.014245664700865746, + "grad_norm": 0.14268416166305542, + "learning_rate": 3.1082609515667424e-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": 0.003939698520116508, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0016180734150111676, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.009497470408678054, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.014245664700865746, + "weighted_rkd_loss": 0.004748735204339027 + }, + { + "epoch": 6.871794871794872, + "flow_matching_loss": 0.014305701572448015, + "grad_norm": 0.22521346807479858, + "learning_rate": 3.048053645644594e-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": 0.005096148257143796, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0019075013406109065, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01209979779087007, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.014305701572448015, + "weighted_rkd_loss": 0.006049898895435035 + }, + { + "epoch": 6.888888888888889, + "flow_matching_loss": 0.013212948385626078, + "grad_norm": 0.148503378033638, + "learning_rate": 2.988369156574916e-06, + "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": 0.003819907363504171, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0015058944351039826, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.0091457090806216, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.013212948385626078, + "weighted_rkd_loss": 0.0045728545403108 + }, + { + "epoch": 6.905982905982906, + "flow_matching_loss": 0.01349337175488472, + "grad_norm": 0.13594625890254974, + "learning_rate": 2.9292100951506225e-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": 0.004378201114013791, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0016772790113463998, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.010433681216090918, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.01349337175488472, + "weighted_rkd_loss": 0.005216840608045459 + }, + { + "epoch": 6.923076923076923, + "flow_matching_loss": 0.013800859311595559, + "grad_norm": 0.17430321872234344, + "learning_rate": 2.8705790491807033e-06, + "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": 0.004294183477759362, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0016260517295449972, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01021441868506372, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.013800859311595559, + "weighted_rkd_loss": 0.00510720934253186 + }, + { + "epoch": 6.94017094017094, + "flow_matching_loss": 0.011465824535116553, + "grad_norm": 0.18166108429431915, + "learning_rate": 2.8124785833770433e-06, + "loss": 0.0165, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.004202205874025822, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0016021716815885156, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.010006583435460925, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.011465824535116553, + "weighted_rkd_loss": 0.005003291717730463 + }, + { + "epoch": 6.957264957264957, + "flow_matching_loss": 0.01547788204625249, + "grad_norm": 0.1667901873588562, + "learning_rate": 2.754911239242241e-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": 0.0038020822452381254, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0018294627370778471, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.009433627175167204, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.01547788204625249, + "weighted_rkd_loss": 0.004716813587583602 + }, + { + "epoch": 6.9743589743589745, + "flow_matching_loss": 0.012793379882350565, + "grad_norm": 0.15985463559627533, + "learning_rate": 2.6978795349584204e-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": 0.004487853916361928, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0018074786174111068, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.010783186368644237, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.012793379882350565, + "weighted_rkd_loss": 0.005391593184322118 + }, + { + "epoch": 6.9914529914529915, + "flow_matching_loss": 0.013774941116571427, + "grad_norm": 0.16853109002113342, + "learning_rate": 2.641385965277088e-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": 0.004574715811759234, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0018272122426424176, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01097664381377399, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.013774941116571427, + "weighted_rkd_loss": 0.005488321906886995 + }, + { + "epoch": 7.0085470085470085, + "flow_matching_loss": 0.01377521250396967, + "grad_norm": 0.16738949716091156, + "learning_rate": 2.5854330014099923e-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": 0.004262465052306652, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0014018778572790324, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.009926808020099997, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.01377521250396967, + "weighted_rkd_loss": 0.004963404010049998 + }, + { + "epoch": 7.0256410256410255, + "flow_matching_loss": 0.013304527848958969, + "grad_norm": 0.1807527244091034, + "learning_rate": 2.5300230909210298e-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": 0.004257907392457128, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0017975858820136637, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.010313400626182556, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.013304527848958969, + "weighted_rkd_loss": 0.005156700313091278 + }, + { + "epoch": 7.042735042735043, + "flow_matching_loss": 0.012193146161735058, + "grad_norm": 0.2061319798231125, + "learning_rate": 2.4751586576191983e-06, + "loss": 0.0171, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.0041587944375351075, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.001565186760853976, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.009882775600999594, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.012193146161735058, + "weighted_rkd_loss": 0.004941387800499797 + }, + { + "epoch": 7.05982905982906, + "flow_matching_loss": 0.01150200036354363, + "grad_norm": 0.12688463926315308, + "learning_rate": 2.420842101452538e-06, + "loss": 0.0163, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.003888703393749893, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0018680165347177535, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.009645423293113709, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.01150200036354363, + "weighted_rkd_loss": 0.004822711646556854 + }, + { + "epoch": 7.076923076923077, + "flow_matching_loss": 0.011715107830241323, + "grad_norm": 0.17974625527858734, + "learning_rate": 2.3670757984031816e-06, + "loss": 0.0162, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.003783612954430282, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0014903668721672147, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.009057592926546931, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.011715107830241323, + "weighted_rkd_loss": 0.0045287964632734655 + }, + { + "epoch": 7.094017094017094, + "flow_matching_loss": 0.011287252325564623, + "grad_norm": 0.1417006403207779, + "learning_rate": 2.313862100383404e-06, + "loss": 0.0154, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.00343161195050925, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0013535286823753267, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.008216752624139189, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.011287252325564623, + "weighted_rkd_loss": 0.0041083763120695945 + }, + { + "epoch": 7.111111111111111, + "flow_matching_loss": 0.012390893045812845, + "grad_norm": 0.1228102594614029, + "learning_rate": 2.2612033351327406e-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": 0.0042154436698183416, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0019080315832979977, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.010338918957859277, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.012390893045812845, + "weighted_rkd_loss": 0.0051694594789296385 + }, + { + "epoch": 7.128205128205128, + "flow_matching_loss": 0.011362318880856037, + "grad_norm": 0.1354866623878479, + "learning_rate": 2.2091018061161788e-06, + "loss": 0.016, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.0038231971440836785, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0017028606845997274, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.009349254984408617, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.011362318880856037, + "weighted_rkd_loss": 0.004674627492204308 + }, + { + "epoch": 7.145299145299146, + "flow_matching_loss": 0.014426226122304797, + "grad_norm": 0.14201338589191437, + "learning_rate": 2.15755979242338e-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": 0.004279016936197877, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.001714422053191811, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.010272455867379904, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.014426226122304797, + "weighted_rkd_loss": 0.005136227933689952 + }, + { + "epoch": 7.162393162393163, + "flow_matching_loss": 0.013876243494451046, + "grad_norm": 0.15545055270195007, + "learning_rate": 2.106579548669e-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": 0.003654197743162513, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0012436681019607932, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.008552063629031182, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.013876243494451046, + "weighted_rkd_loss": 0.004276031814515591 + }, + { + "epoch": 7.17948717948718, + "flow_matching_loss": 0.012172015756368637, + "grad_norm": 0.18450547754764557, + "learning_rate": 2.0561633048940477e-06, + "loss": 0.0172, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.004115495644509792, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0017840529326349497, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.010015044175088406, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.012172015756368637, + "weighted_rkd_loss": 0.005007522087544203 + }, + { + "epoch": 7.196581196581197, + "flow_matching_loss": 0.01326232748106122, + "grad_norm": 0.1385352909564972, + "learning_rate": 2.006313266468355e-06, + "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": 0.004702828754670918, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0018367817858234048, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.011242439271882176, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.01326232748106122, + "weighted_rkd_loss": 0.005621219635941088 + }, + { + "epoch": 7.213675213675214, + "flow_matching_loss": 0.01336014000698924, + "grad_norm": 0.16189557313919067, + "learning_rate": 1.957031613994097e-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": 0.003861038712784648, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0015294211858417838, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.009251498710364103, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.01336014000698924, + "weighted_rkd_loss": 0.0046257493551820515 + }, + { + "epoch": 7.230769230769231, + "flow_matching_loss": 0.013220749609172345, + "grad_norm": 0.18385490775108337, + "learning_rate": 1.9083205032103995e-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": 0.0037725105416029693, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0013661614037118852, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.008911182498559355, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.013220749609172345, + "weighted_rkd_loss": 0.004455591249279678 + }, + { + "epoch": 7.247863247863248, + "flow_matching_loss": 0.01324486918747425, + "grad_norm": 0.18078869581222534, + "learning_rate": 1.8601820648990536e-06, + "loss": 0.0179, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.0039695461047813295, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0013943981553893535, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.009333490440621973, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.01324486918747425, + "weighted_rkd_loss": 0.004666745220310986 + }, + { + "epoch": 7.264957264957265, + "flow_matching_loss": 0.01174166426062584, + "grad_norm": 0.14929896593093872, + "learning_rate": 1.8126184047912936e-06, + "loss": 0.0164, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.0038963052444159984, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0016135171288624405, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.009406127594411373, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.01174166426062584, + "weighted_rkd_loss": 0.004703063797205686 + }, + { + "epoch": 7.282051282051282, + "flow_matching_loss": 0.01100527481175959, + "grad_norm": 0.2219940721988678, + "learning_rate": 1.7656316034757024e-06, + "loss": 0.0167, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.004765593982301653, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0019187520316336303, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.011449940036982298, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.01100527481175959, + "weighted_rkd_loss": 0.005724970018491149 + }, + { + "epoch": 7.299145299145299, + "flow_matching_loss": 0.013349955342710019, + "grad_norm": 0.12957392632961273, + "learning_rate": 1.7192237163071816e-06, + "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": 0.003737821336835623, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.001505720429122448, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.008981363102793693, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.013349955342710019, + "weighted_rkd_loss": 0.004490681551396846 + }, + { + "epoch": 7.316239316239316, + "flow_matching_loss": 0.012917829304933548, + "grad_norm": 0.13917148113250732, + "learning_rate": 1.6733967733170573e-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": 0.0037979158572852612, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.002031996531877667, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.009627828234806656, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.012917829304933548, + "weighted_rkd_loss": 0.004813914117403328 + }, + { + "epoch": 7.333333333333333, + "flow_matching_loss": 0.011703525995835662, + "grad_norm": 0.19382449984550476, + "learning_rate": 1.628152779124279e-06, + "loss": 0.0166, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.004080761969089508, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.001575100829359144, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.009736624732613563, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.011703525995835662, + "weighted_rkd_loss": 0.004868312366306781 + }, + { + "epoch": 7.35042735042735, + "flow_matching_loss": 0.01296322252601385, + "grad_norm": 0.14509230852127075, + "learning_rate": 1.583493712847715e-06, + "loss": 0.0174, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.0037901628063991668, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0013702258991543203, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.008950551599264145, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.01296322252601385, + "weighted_rkd_loss": 0.004475275799632072 + }, + { + "epoch": 7.367521367521368, + "flow_matching_loss": 0.014123854786157608, + "grad_norm": 0.1688782125711441, + "learning_rate": 1.5394215280196084e-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": 0.0036876884289085863, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0015890191541984677, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.008964395988732577, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.014123854786157608, + "weighted_rkd_loss": 0.004482197994366288 + }, + { + "epoch": 7.384615384615385, + "flow_matching_loss": 0.013744353968650102, + "grad_norm": 0.23357872664928436, + "learning_rate": 1.4959381525000904e-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": 0.0046668478287756445, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0017815333674661815, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01111522908322513, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.013744353968650102, + "weighted_rkd_loss": 0.005557614541612565 + }, + { + "epoch": 7.401709401709402, + "flow_matching_loss": 0.013115393370389939, + "grad_norm": 0.13692398369312286, + "learning_rate": 1.453045488392879e-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": 0.004141918360255658, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.001547129702521488, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.009830966452136635, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.013115393370389939, + "weighted_rkd_loss": 0.004915483226068318 + }, + { + "epoch": 7.418803418803419, + "flow_matching_loss": 0.012881484348326921, + "grad_norm": 0.20485293865203857, + "learning_rate": 1.4107454119620484e-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": 0.003988063824363053, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0015699075884185731, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.009546035202220082, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.012881484348326921, + "weighted_rkd_loss": 0.004773017601110041 + }, + { + "epoch": 7.435897435897436, + "flow_matching_loss": 0.01345738205127418, + "grad_norm": 0.2101745456457138, + "learning_rate": 1.3690397735499778e-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": 0.003833774570375681, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0017207548720762134, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.009388304036110639, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.01345738205127418, + "weighted_rkd_loss": 0.0046941520180553194 + }, + { + "epoch": 7.452991452991453, + "flow_matching_loss": 0.014391318708658219, + "grad_norm": 0.17784802615642548, + "learning_rate": 1.327930397496403e-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": 0.004032033658586442, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.001528675900772214, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.009592743217945099, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.014391318708658219, + "weighted_rkd_loss": 0.004796371608972549 + }, + { + "epoch": 7.47008547008547, + "flow_matching_loss": 0.011513168923556805, + "grad_norm": 0.18674953281879425, + "learning_rate": 1.287419082058604e-06, + "loss": 0.0159, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.0035867713391780852, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0016056925931479781, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.008779235230758787, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.011513168923556805, + "weighted_rkd_loss": 0.0043896176153793934 + }, + { + "epoch": 7.487179487179487, + "flow_matching_loss": 0.012771233264356852, + "grad_norm": 0.16635802388191223, + "learning_rate": 1.2475075993327589e-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": 0.004345070035196841, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.00172021365724504, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.010410353634506464, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.012771233264356852, + "weighted_rkd_loss": 0.005205176817253232 + }, + { + "epoch": 7.504273504273504, + "flow_matching_loss": 0.013028046255931259, + "grad_norm": 0.11585721373558044, + "learning_rate": 1.208197695176413e-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": 0.004318495141342282, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0016846387938130648, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.010321629187092184, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.013028046255931259, + "weighted_rkd_loss": 0.005160814593546092 + }, + { + "epoch": 7.521367521367521, + "flow_matching_loss": 0.011222801310941577, + "grad_norm": 0.14337489008903503, + "learning_rate": 1.1694910891321259e-06, + "loss": 0.016, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.003986520576290786, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0016476838150992989, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.009620724944397807, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.011222801310941577, + "weighted_rkd_loss": 0.0048103624721989036 + }, + { + "epoch": 7.538461538461538, + "flow_matching_loss": 0.014170465990900993, + "grad_norm": 0.1666136533021927, + "learning_rate": 1.1313894743522363e-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": 0.004018307896330953, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0017779929097741843, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.00981460870243609, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.014170465990900993, + "weighted_rkd_loss": 0.004907304351218045 + }, + { + "epoch": 7.555555555555555, + "flow_matching_loss": 0.012973797554150224, + "grad_norm": 0.1495077759027481, + "learning_rate": 1.0938945175248016e-06, + "loss": 0.0172, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.00348243482876569, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.001476709794951603, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.008441579481586813, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.012973797554150224, + "weighted_rkd_loss": 0.004220789740793407 + }, + { + "epoch": 7.572649572649572, + "flow_matching_loss": 0.013660789560526609, + "grad_norm": 0.1573425829410553, + "learning_rate": 1.0570078588007086e-06, + "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": 0.0042850013822317125, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0018650735961273312, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.010435076337307692, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.013660789560526609, + "weighted_rkd_loss": 0.005217538168653846 + }, + { + "epoch": 7.589743589743589, + "flow_matching_loss": 0.010968834720551967, + "grad_norm": 0.17911341786384583, + "learning_rate": 1.0207311117219009e-06, + "loss": 0.0156, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.0039019836112856865, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0014731241972185671, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.009277091501280666, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.010968834720551967, + "weighted_rkd_loss": 0.004638545750640333 + }, + { + "epoch": 7.6068376068376065, + "flow_matching_loss": 0.012092063948512078, + "grad_norm": 0.21830224990844727, + "learning_rate": 9.850658631508198e-07, + "loss": 0.0172, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.004266565758734941, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0017167957383207976, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.010249927174299955, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.012092063948512078, + "weighted_rkd_loss": 0.005124963587149978 + }, + { + "epoch": 7.6239316239316235, + "flow_matching_loss": 0.013988563138991595, + "grad_norm": 0.14057599008083344, + "learning_rate": 9.500136732009751e-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": 0.0038496993482112886, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0015126889105886222, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.009212087607011199, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.013988563138991595, + "weighted_rkd_loss": 0.0046060438035055995 + }, + { + "epoch": 7.641025641025641, + "flow_matching_loss": 0.010955549450591207, + "grad_norm": 0.18209223449230194, + "learning_rate": 9.155760751687054e-07, + "loss": 0.0157, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.003972943243570626, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0015911909751594067, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.009537077462300658, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.010955549450591207, + "weighted_rkd_loss": 0.004768538731150329 + }, + { + "epoch": 7.6581196581196584, + "flow_matching_loss": 0.012481959071010351, + "grad_norm": 0.21544641256332397, + "learning_rate": 8.817545754661155e-07, + "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": 0.004292281693778932, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0016132907243445516, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01019785408861935, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.012481959071010351, + "weighted_rkd_loss": 0.005098927044309675 + }, + { + "epoch": 7.6752136752136755, + "flow_matching_loss": 0.014317840617150069, + "grad_norm": 0.13255079090595245, + "learning_rate": 8.485506535551668e-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": 0.0034783784998580814, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0013597151497378944, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.008316472126170992, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.014317840617150069, + "weighted_rkd_loss": 0.004158236063085496 + }, + { + "epoch": 7.6923076923076925, + "flow_matching_loss": 0.010459744790568947, + "grad_norm": 0.17515155673027039, + "learning_rate": 8.159657618829685e-07, + "loss": 0.0152, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.004000649321824312, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0014889903599396349, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.009490288980305194, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.010459744790568947, + "weighted_rkd_loss": 0.004745144490152597 + }, + { + "epoch": 7.7094017094017095, + "flow_matching_loss": 0.01422869712114334, + "grad_norm": 0.17458616197109222, + "learning_rate": 7.840013258182471e-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": 0.003670058771967888, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.001456214237259701, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.008796331845223904, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.01422869712114334, + "weighted_rkd_loss": 0.004398165922611952 + }, + { + "epoch": 7.726495726495727, + "flow_matching_loss": 0.013734146859496832, + "grad_norm": 0.14348195493221283, + "learning_rate": 7.526587435889809e-07, + "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": 0.003523202263750136, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.001518395874882117, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.008564800256863236, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.013734146859496832, + "weighted_rkd_loss": 0.004282400128431618 + }, + { + "epoch": 7.743589743589744, + "flow_matching_loss": 0.01234424621798098, + "grad_norm": 0.1296064853668213, + "learning_rate": 7.219393862212593e-07, + "loss": 0.0167, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.003714152844622731, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0013201057794503868, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.008748411480337382, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.01234424621798098, + "weighted_rkd_loss": 0.004374205740168691 + }, + { + "epoch": 7.760683760683761, + "flow_matching_loss": 0.012772375950589776, + "grad_norm": 0.1534041166305542, + "learning_rate": 6.918445974792819e-07, + "loss": 0.0165, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.0030991335166618227, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0011701180832460524, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.007368385139852762, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.012772375950589776, + "weighted_rkd_loss": 0.003684192569926381 + }, + { + "epoch": 7.777777777777778, + "flow_matching_loss": 0.012171571562066674, + "grad_norm": 0.10533745586872101, + "learning_rate": 6.6237569380661e-07, + "loss": 0.0166, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.003707205457612872, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0014780213998164982, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.008892432367429138, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.012171571562066674, + "weighted_rkd_loss": 0.004446216183714569 + }, + { + "epoch": 7.794871794871795, + "flow_matching_loss": 0.012081097718328238, + "grad_norm": 0.11634702980518341, + "learning_rate": 6.335339642685495e-07, + "loss": 0.0163, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.0034490347607061267, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.001452084380434826, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.008350153779610991, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.012081097718328238, + "weighted_rkd_loss": 0.0041750768898054956 + }, + { + "epoch": 7.811965811965812, + "flow_matching_loss": 0.013012279709801077, + "grad_norm": 0.18919211626052856, + "learning_rate": 6.05320670495787e-07, + "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": 0.004376333625987172, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0016027019592002034, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.010355369048193097, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.013012279709801077, + "weighted_rkd_loss": 0.005177684524096548 + }, + { + "epoch": 7.829059829059829, + "flow_matching_loss": 0.011991818202659488, + "grad_norm": 0.17841693758964539, + "learning_rate": 5.777370466291954e-07, + "loss": 0.0166, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.0039478613995015625, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0013961193850263952, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.00929184230044484, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.011991818202659488, + "weighted_rkd_loss": 0.00464592115022242 + }, + { + "epoch": 7.846153846153846, + "flow_matching_loss": 0.010694775450974702, + "grad_norm": 0.15433743596076965, + "learning_rate": 5.507842992658336e-07, + "loss": 0.0148, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.0033815303584560754, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0013989461411256343, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.008162006922066211, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.010694775450974702, + "weighted_rkd_loss": 0.0040810034610331055 + }, + { + "epoch": 7.863247863247864, + "flow_matching_loss": 0.014666303992271423, + "grad_norm": 0.1808875948190689, + "learning_rate": 5.24463607406192e-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": 0.00379934289958328, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0016835816204547882, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.009282267419621348, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.014666303992271423, + "weighted_rkd_loss": 0.004641133709810674 + }, + { + "epoch": 7.880341880341881, + "flow_matching_loss": 0.011279391311109065, + "grad_norm": 0.17627708613872528, + "learning_rate": 4.987761224025944e-07, + "loss": 0.0163, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.004237288725562394, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0014775051502510905, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.009952082624658942, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.011279391311109065, + "weighted_rkd_loss": 0.004976041312329471 + }, + { + "epoch": 7.897435897435898, + "flow_matching_loss": 0.015158367482945323, + "grad_norm": 0.14843781292438507, + "learning_rate": 4.737229679088556e-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": 0.004200689448043704, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.00160459796898067, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01000597677193582, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.015158367482945323, + "weighted_rkd_loss": 0.00500298838596791 + }, + { + "epoch": 7.914529914529915, + "flow_matching_loss": 0.013612598646432161, + "grad_norm": 0.1926652193069458, + "learning_rate": 4.493052398311115e-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": 0.0039387172088027, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0016255067079328001, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.009502941137179732, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.013612598646432161, + "weighted_rkd_loss": 0.004751470568589866 + }, + { + "epoch": 7.931623931623932, + "flow_matching_loss": 0.0120464526116848, + "grad_norm": 0.16931195557117462, + "learning_rate": 4.2552400627989043e-07, + "loss": 0.0168, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.004019210627302527, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0015675419068429618, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.009605963109061122, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.0120464526116848, + "weighted_rkd_loss": 0.004802981554530561 + }, + { + "epoch": 7.948717948717949, + "flow_matching_loss": 0.012973124394193292, + "grad_norm": 0.1812753826379776, + "learning_rate": 4.023803075233923e-07, + "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": 0.004355783481150866, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0017748346959706395, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.010486401710659265, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.012973124394193292, + "weighted_rkd_loss": 0.005243200855329633 + }, + { + "epoch": 7.965811965811966, + "flow_matching_loss": 0.013465069141238928, + "grad_norm": 0.16017380356788635, + "learning_rate": 3.7987515594196997e-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": 0.0038325405679643155, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.001507552177645266, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.009172633336856962, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.013465069141238928, + "weighted_rkd_loss": 0.004586316668428481 + }, + { + "epoch": 7.982905982905983, + "flow_matching_loss": 0.012850679084658622, + "grad_norm": 0.14726558327674866, + "learning_rate": 3.580095359838648e-07, + "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": 0.004245123150758446, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0017716314876452088, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.010261877812445164, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.012850679084658622, + "weighted_rkd_loss": 0.005130938906222582 + }, + { + "epoch": 8.0, + "flow_matching_loss": 0.013819062430411578, + "grad_norm": 0.14340880513191223, + "learning_rate": 3.3678440412212576e-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": 0.00434499008115381, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0017850894946604967, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01047506956383586, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.013819062430411578, + "weighted_rkd_loss": 0.00523753478191793 + }, + { + "epoch": 8.017094017094017, + "flow_matching_loss": 0.014346749149262905, + "grad_norm": 0.1989343911409378, + "learning_rate": 3.162006888127833e-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": 0.0036725772311910986, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.001292025181464851, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.008637179713696241, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.014346749149262905, + "weighted_rkd_loss": 0.0043185898568481205 + }, + { + "epoch": 8.034188034188034, + "flow_matching_loss": 0.01425888082012534, + "grad_norm": 0.1949833631515503, + "learning_rate": 2.9625929045422327e-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": 0.003957598004490137, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.001433904911391437, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.009349100943654775, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.01425888082012534, + "weighted_rkd_loss": 0.0046745504718273875 + }, + { + "epoch": 8.051282051282051, + "flow_matching_loss": 0.013923675566911698, + "grad_norm": 0.16496892273426056, + "learning_rate": 2.7696108134781186e-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": 0.00467620249837637, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0017530645069200545, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.011105469428002834, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.013923675566911698, + "weighted_rkd_loss": 0.005552734714001417 + }, + { + "epoch": 8.068376068376068, + "flow_matching_loss": 0.013424544688314199, + "grad_norm": 0.1507738083600998, + "learning_rate": 2.583069056597376e-07, + "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": 0.0038854417391121386, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0017710975953377783, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.009541981061920524, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.013424544688314199, + "weighted_rkd_loss": 0.004770990530960262 + }, + { + "epoch": 8.085470085470085, + "flow_matching_loss": 0.011561490455642342, + "grad_norm": 0.16206304728984833, + "learning_rate": 2.4029757938407116e-07, + "loss": 0.0165, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.004052890441380441, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0017420763499103487, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.009847857290878892, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.011561490455642342, + "weighted_rkd_loss": 0.004923928645439446 + }, + { + "epoch": 8.102564102564102, + "flow_matching_loss": 0.01580977626144886, + "grad_norm": 0.20269455015659332, + "learning_rate": 2.229338903070871e-07, + "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": 0.003932159813120961, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0016555070236790926, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.009519826667383314, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.01580977626144886, + "weighted_rkd_loss": 0.004759913333691657 + }, + { + "epoch": 8.11965811965812, + "flow_matching_loss": 0.013072844501584769, + "grad_norm": 0.19857940077781677, + "learning_rate": 2.0621659797279645e-07, + "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": 0.0038817472057417037, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.001545777020510286, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.009309271443635225, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.013072844501584769, + "weighted_rkd_loss": 0.0046546357218176125 + }, + { + "epoch": 8.136752136752136, + "flow_matching_loss": 0.013049683813005687, + "grad_norm": 0.12958791851997375, + "learning_rate": 1.901464336497233e-07, + "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": 0.00369636001996696, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0015062186168506742, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.008898938726633786, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.013049683813005687, + "weighted_rkd_loss": 0.004449469363316893 + }, + { + "epoch": 8.153846153846153, + "flow_matching_loss": 0.013017881475389004, + "grad_norm": 0.16941314935684204, + "learning_rate": 1.7472410029891527e-07, + "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": 0.0036582418833859265, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.001561993802897632, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.008878477476537227, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.013017881475389004, + "weighted_rkd_loss": 0.004439238738268614 + }, + { + "epoch": 8.17094017094017, + "flow_matching_loss": 0.012620376516133546, + "grad_norm": 0.12690627574920654, + "learning_rate": 1.5995027254319173e-07, + "loss": 0.0167, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.003393899230286479, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0014338911918457598, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.008221689658239485, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.012620376516133546, + "weighted_rkd_loss": 0.004110844829119742 + }, + { + "epoch": 8.188034188034187, + "flow_matching_loss": 0.011777253774926066, + "grad_norm": 0.11281920969486237, + "learning_rate": 1.4582559663764372e-07, + "loss": 0.0164, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.003883430501446128, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0015375308750662953, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.009304391825571657, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.011777253774926066, + "weighted_rkd_loss": 0.004652195912785828 + }, + { + "epoch": 8.205128205128204, + "flow_matching_loss": 0.013854224048554898, + "grad_norm": 0.1607847958803177, + "learning_rate": 1.323506904413485e-07, + "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": 0.0034347342560067775, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0014547378872521223, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.008324206387624145, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.013854224048554898, + "weighted_rkd_loss": 0.004162103193812073 + }, + { + "epoch": 8.222222222222221, + "flow_matching_loss": 0.011663326108828188, + "grad_norm": 0.16224442422389984, + "learning_rate": 1.195261433903594e-07, + "loss": 0.0165, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.004068200034089387, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0015405308920890092, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.009676930774003268, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.011663326108828188, + "weighted_rkd_loss": 0.004838465387001634 + }, + { + "epoch": 8.239316239316238, + "flow_matching_loss": 0.011585106793791055, + "grad_norm": 0.1953214555978775, + "learning_rate": 1.0735251647190969e-07, + "loss": 0.0164, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.004023355850949884, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0015609974274411798, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.009607709199190139, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.011585106793791055, + "weighted_rkd_loss": 0.0048038545995950695 + }, + { + "epoch": 8.256410256410255, + "flow_matching_loss": 0.010779148479923607, + "grad_norm": 0.18219971656799316, + "learning_rate": 9.583034219987408e-08, + "loss": 0.0152, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.003688325686380267, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0014612552186008543, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.00883790673688054, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.010779148479923607, + "weighted_rkd_loss": 0.00441895336844027 + }, + { + "epoch": 8.273504273504274, + "flow_matching_loss": 0.01083608423359692, + "grad_norm": 0.163270503282547, + "learning_rate": 8.496012459148717e-08, + "loss": 0.0152, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.003690268355421722, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0013255010242573916, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.008706037746742367, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.01083608423359692, + "weighted_rkd_loss": 0.004353018873371184 + }, + { + "epoch": 8.290598290598291, + "flow_matching_loss": 0.014668568409979344, + "grad_norm": 0.1465245932340622, + "learning_rate": 7.474233914527618e-08, + "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": 0.0032832271186634897, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0012717477569822223, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.00783820180222392, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.014668568409979344, + "weighted_rkd_loss": 0.00391910090111196 + }, + { + "epoch": 8.307692307692308, + "flow_matching_loss": 0.014427309110760688, + "grad_norm": 0.17697834968566895, + "learning_rate": 6.517743282028254e-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": 0.004399512312375009, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0016026621451601386, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.010401686839759349, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.014427309110760688, + "weighted_rkd_loss": 0.0052008434198796746 + }, + { + "epoch": 8.324786324786325, + "flow_matching_loss": 0.0124134442768991, + "grad_norm": 0.16188935935497284, + "learning_rate": 5.6265824016492604e-08, + "loss": 0.0174, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.004069513431750238, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0019054210744798184, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.010044447798281907, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.0124134442768991, + "weighted_rkd_loss": 0.005022223899140954 + }, + { + "epoch": 8.341880341880342, + "flow_matching_loss": 0.0162894481793046, + "grad_norm": 0.2575010061264038, + "learning_rate": 4.800790255654719e-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": 0.004137730854563415, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0018405434326268732, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.010116005130112172, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.0162894481793046, + "weighted_rkd_loss": 0.005058002565056086 + }, + { + "epoch": 8.35897435897436, + "flow_matching_loss": 0.014363493211567402, + "grad_norm": 0.1337166279554367, + "learning_rate": 4.040402966867862e-08, + "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": 0.004561483091674745, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0017737012123689056, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01089666741900146, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.014363493211567402, + "weighted_rkd_loss": 0.00544833370950073 + }, + { + "epoch": 8.376068376068377, + "flow_matching_loss": 0.012774949660524726, + "grad_norm": 0.1732391119003296, + "learning_rate": 3.3454537970925014e-08, + "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": 0.004006552579812706, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0016211368376389146, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.00963424202054739, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.012774949660524726, + "weighted_rkd_loss": 0.004817121010273695 + }, + { + "epoch": 8.393162393162394, + "flow_matching_loss": 0.015056971274316312, + "grad_norm": 0.14762836694717407, + "learning_rate": 2.715973145656192e-08, + "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": 0.0035704671638086436, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0013850737072061748, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.008526008063927292, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.015056971274316312, + "weighted_rkd_loss": 0.004263004031963646 + }, + { + "epoch": 8.41025641025641, + "flow_matching_loss": 0.01169312335550785, + "grad_norm": 0.1522114872932434, + "learning_rate": 2.151988548082129e-08, + "loss": 0.0162, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.0037407787516713143, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.001625890377908945, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.009107447927817702, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.01169312335550785, + "weighted_rkd_loss": 0.004553723963908851 + }, + { + "epoch": 8.427350427350428, + "flow_matching_loss": 0.012849413696676493, + "grad_norm": 0.14591342210769653, + "learning_rate": 1.6535246748837772e-08, + "loss": 0.0174, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.003902881685644388, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0013816383783705533, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.009187401831150055, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.012849413696676493, + "weighted_rkd_loss": 0.004593700915575027 + }, + { + "epoch": 8.444444444444445, + "flow_matching_loss": 0.013146649487316609, + "grad_norm": 0.15642616152763367, + "learning_rate": 1.2206033304855702e-08, + "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": 0.003651187289506197, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0015399788680952042, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.008842353383079172, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.013146649487316609, + "weighted_rkd_loss": 0.004421176691539586 + }, + { + "epoch": 8.461538461538462, + "flow_matching_loss": 0.013237102329730988, + "grad_norm": 0.17536711692810059, + "learning_rate": 8.532434522695031e-09, + "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": 0.004425908089615405, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.001877948606852442, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.010729764867573977, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.013237102329730988, + "weighted_rkd_loss": 0.005364882433786989 + }, + { + "epoch": 8.478632478632479, + "flow_matching_loss": 0.013877835730090738, + "grad_norm": 0.1576084941625595, + "learning_rate": 5.5146110974679805e-09, + "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": 0.004351154412142932, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.001673866022611037, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.010376174934208393, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.013877835730090738, + "weighted_rkd_loss": 0.0051880874671041965 + }, + { + "epoch": 8.495726495726496, + "flow_matching_loss": 0.012626701407134533, + "grad_norm": 0.19507397711277008, + "learning_rate": 3.1526950385496467e-09, + "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": 0.004268162348307669, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0017047684756107627, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.010241093207150697, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.012626701407134533, + "weighted_rkd_loss": 0.005120546603575349 + }, + { + "epoch": 8.512820512820513, + "flow_matching_loss": 0.010833437461405993, + "grad_norm": 0.20359574258327484, + "learning_rate": 1.4467896637959666e-09, + "loss": 0.0159, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.0043221628991886975, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0015077884832862764, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.010152114275842904, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.010833437461405993, + "weighted_rkd_loss": 0.005076057137921452 + }, + { + "epoch": 8.52991452991453, + "flow_matching_loss": 0.013263375498354435, + "grad_norm": 0.15500546991825104, + "learning_rate": 3.9696959503898733e-10, + "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": 0.003604265209287405, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0013148864556569606, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.008523416798561811, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.013263375498354435, + "weighted_rkd_loss": 0.0042617083992809056 + }, + { + "epoch": 8.547008547008547, + "flow_matching_loss": 0.01256059342995286, + "grad_norm": 0.2574502229690552, + "learning_rate": 3.280754806156416e-12, + "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": 0.004128366347867995, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.001785930647747591, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.010042663337662816, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.01256059342995286, + "weighted_rkd_loss": 0.005021331668831408 + } + ], + "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/seed2/default/train_5shot_seed2_fullactionhead_5k/config.json b/FewShot/39tasks/5shot/FromRKD-Timewarp_FullactionHead/seed2/default/train_5shot_seed2_fullactionhead_5k/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/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/model-00001-of-00002.safetensors b/FewShot/39tasks/5shot/FromRKD-Timewarp_FullactionHead/seed2/default/train_5shot_seed2_fullactionhead_5k/model-00001-of-00002.safetensors new file mode 100644 index 0000000000000000000000000000000000000000..c90f52dd173a01f62cc39f7ac3fcce55ca525157 --- /dev/null +++ b/FewShot/39tasks/5shot/FromRKD-Timewarp_FullactionHead/seed2/default/train_5shot_seed2_fullactionhead_5k/model-00001-of-00002.safetensors @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:77aaacc7ee85470538018578fecaded4f70641bd8f27d08ebee25a9a605e7a99 +size 4999367032 diff --git a/FewShot/39tasks/5shot/FromRKD-Timewarp_FullactionHead/seed2/default/train_5shot_seed2_fullactionhead_5k/model-00002-of-00002.safetensors b/FewShot/39tasks/5shot/FromRKD-Timewarp_FullactionHead/seed2/default/train_5shot_seed2_fullactionhead_5k/model-00002-of-00002.safetensors new file mode 100644 index 0000000000000000000000000000000000000000..40017848604c545805a73648b52df0f0d03a2059 --- /dev/null +++ b/FewShot/39tasks/5shot/FromRKD-Timewarp_FullactionHead/seed2/default/train_5shot_seed2_fullactionhead_5k/model-00002-of-00002.safetensors @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:e239dee05215eab5d4798f2b6cf01904a026fd99af2c3f053cabc4a4b3f41d74 +size 2664540348 diff --git a/FewShot/39tasks/5shot/FromRKD-Timewarp_FullactionHead/seed2/default/train_5shot_seed2_fullactionhead_5k/model.safetensors.index.json b/FewShot/39tasks/5shot/FromRKD-Timewarp_FullactionHead/seed2/default/train_5shot_seed2_fullactionhead_5k/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/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/trainer_state.json b/FewShot/39tasks/5shot/FromRKD-Timewarp_FullactionHead/seed2/default/train_5shot_seed2_fullactionhead_5k/trainer_state.json new file mode 100644 index 0000000000000000000000000000000000000000..be5fa0c71b731686f65c0b27a4879713eb7afbde --- /dev/null +++ b/FewShot/39tasks/5shot/FromRKD-Timewarp_FullactionHead/seed2/default/train_5shot_seed2_fullactionhead_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": 0.1910901501774788, + "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 + }, + { + "epoch": 5.145299145299146, + "flow_matching_loss": 0.015014776308089495, + "grad_norm": 0.18950475752353668, + "learning_rate": 1.1231228239056467e-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": 0.005329322838224471, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0022741535911336543, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.012932799477130174, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.015014776308089495, + "weighted_rkd_loss": 0.006466399738565087 + }, + { + "epoch": 5.162393162393163, + "flow_matching_loss": 0.015901549346745016, + "grad_norm": 0.14783582091331482, + "learning_rate": 1.113528559434845e-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": 0.005319552915170789, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.002164692198857665, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.012803797982633114, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.015901549346745016, + "weighted_rkd_loss": 0.006401898991316557 + }, + { + "epoch": 5.17948717948718, + "flow_matching_loss": 0.016365189291536807, + "grad_norm": 0.15554279088974, + "learning_rate": 1.1039512004768476e-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": 0.004656393500044942, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.002018486650194973, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.011331273661926388, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.016365189291536807, + "weighted_rkd_loss": 0.005665636830963194 + }, + { + "epoch": 5.196581196581197, + "flow_matching_loss": 0.01562155494466424, + "grad_norm": 0.19159680604934692, + "learning_rate": 1.0943911659763612e-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": 0.005035868706181645, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.002107031480409205, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.012178768776357174, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.01562155494466424, + "weighted_rkd_loss": 0.006089384388178587 + }, + { + "epoch": 5.213675213675214, + "flow_matching_loss": 0.015252807457000018, + "grad_norm": 0.1729981005191803, + "learning_rate": 1.084848874120266e-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": 0.0056610606145113705, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0025155096664093435, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.013837630953639745, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.015252807457000018, + "weighted_rkd_loss": 0.006918815476819873 + }, + { + "epoch": 5.230769230769231, + "flow_matching_loss": 0.017009658738970757, + "grad_norm": 0.20856738090515137, + "learning_rate": 1.0753247423193196e-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": 0.005203615548089147, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0021950440714135766, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.012602275051176549, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.017009658738970757, + "weighted_rkd_loss": 0.006301137525588274 + }, + { + "epoch": 5.247863247863248, + "flow_matching_loss": 0.014320117607712745, + "grad_norm": 0.200138121843338, + "learning_rate": 1.0658191871899012e-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": 0.004975240374915302, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0019149117753840983, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.011865392327308655, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.014320117607712745, + "weighted_rkd_loss": 0.005932696163654327 + }, + { + "epoch": 5.264957264957265, + "flow_matching_loss": 0.01623127833008766, + "grad_norm": 0.1671903431415558, + "learning_rate": 1.0563326245357852e-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": 0.005758790764957666, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.002532364521175623, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.014049946237355471, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.01623127833008766, + "weighted_rkd_loss": 0.0070249731186777355 + }, + { + "epoch": 5.282051282051282, + "flow_matching_loss": 0.013452016282826662, + "grad_norm": 0.1713135540485382, + "learning_rate": 1.0468654693299557e-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": 0.0055105118080973625, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0022663405281491576, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01328736413270235, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.013452016282826662, + "weighted_rkd_loss": 0.006643682066351175 + }, + { + "epoch": 5.299145299145299, + "flow_matching_loss": 0.015398501604795455, + "grad_norm": 0.15591202676296234, + "learning_rate": 1.0374181356964501e-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": 0.005339693161658943, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.002155479183420539, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.012834865320473909, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.015398501604795455, + "weighted_rkd_loss": 0.006417432660236954 + }, + { + "epoch": 5.316239316239316, + "flow_matching_loss": 0.013474423624575137, + "grad_norm": 0.186759814620018, + "learning_rate": 1.0279910368922465e-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": 0.005282991589047015, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.00228860656498, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.012854589708149433, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.013474423624575137, + "weighted_rkd_loss": 0.006427294854074716 + }, + { + "epoch": 5.333333333333333, + "flow_matching_loss": 0.014557658508419991, + "grad_norm": 0.19690386950969696, + "learning_rate": 1.018584585289186e-05, + "loss": 0.021, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.005379770835861564, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0021475294954143463, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01290707103908062, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.014557658508419991, + "weighted_rkd_loss": 0.00645353551954031 + }, + { + "epoch": 5.35042735042735, + "flow_matching_loss": 0.016512978728860615, + "grad_norm": 0.24174918234348297, + "learning_rate": 1.0091991923559348e-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": 0.005569035932421684, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0023933482705615463, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.013531420193612576, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.016512978728860615, + "weighted_rkd_loss": 0.006765710096806288 + }, + { + "epoch": 5.367521367521368, + "flow_matching_loss": 0.015567683707922697, + "grad_norm": 0.17910248041152954, + "learning_rate": 9.99835268639984e-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": 0.005675594275817275, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0023775284411385656, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.013728717155754567, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.015567683707922697, + "weighted_rkd_loss": 0.0068643585778772834 + }, + { + "epoch": 5.384615384615385, + "flow_matching_loss": 0.012863228842616082, + "grad_norm": 0.17725598812103271, + "learning_rate": 9.904932237496913e-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": 0.004814130300655961, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.002020637388341129, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.011648897919803858, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.012863228842616082, + "weighted_rkd_loss": 0.005824448959901929 + }, + { + "epoch": 5.401709401709402, + "flow_matching_loss": 0.012962156720459462, + "grad_norm": 0.21131287515163422, + "learning_rate": 9.811734663363659e-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": 0.005019215657375753, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0023267841315828264, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.012365215551108122, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.012962156720459462, + "weighted_rkd_loss": 0.006182607775554061 + }, + { + "epoch": 5.418803418803419, + "flow_matching_loss": 0.016335158282890914, + "grad_norm": 0.16600927710533142, + "learning_rate": 9.718764040763875e-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": 0.005628416361287236, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0027493985020555557, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.014006231166422366, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.016335158282890914, + "weighted_rkd_loss": 0.007003115583211183 + }, + { + "epoch": 5.435897435897436, + "flow_matching_loss": 0.012864034436643124, + "grad_norm": 0.2108919471502304, + "learning_rate": 9.626024436533777e-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": 0.0059022421948611735, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0023489799350500107, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.014153464417904615, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.012864034436643124, + "weighted_rkd_loss": 0.007076732208952307 + }, + { + "epoch": 5.452991452991453, + "flow_matching_loss": 0.014886892028152942, + "grad_norm": 0.21139591932296753, + "learning_rate": 9.533519907404095e-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": 0.005289020924828946, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.002200934721622616, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.012778976373374462, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.014886892028152942, + "weighted_rkd_loss": 0.006389488186687231 + }, + { + "epoch": 5.47008547008547, + "flow_matching_loss": 0.016243355721235274, + "grad_norm": 0.13592956960201263, + "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": 0.005620071780867875, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.002551423793192953, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.013791567226871849, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.016243355721235274, + "weighted_rkd_loss": 0.006895783613435924 + }, + { + "epoch": 5.487179487179487, + "flow_matching_loss": 0.015402343589812517, + "grad_norm": 0.14428843557834625, + "learning_rate": 9.349232249777124e-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": 0.005076029896736145, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.002067275461740792, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.012219335231930018, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.015402343589812517, + "weighted_rkd_loss": 0.006109667615965009 + }, + { + "epoch": 5.504273504273504, + "flow_matching_loss": 0.01771446131169796, + "grad_norm": 0.16076010465621948, + "learning_rate": 9.257457182618985e-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": 0.005623388360254467, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0022086908458732067, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.013455467531457543, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.01771446131169796, + "weighted_rkd_loss": 0.006727733765728772 + }, + { + "epoch": 5.521367521367521, + "flow_matching_loss": 0.014067314937710762, + "grad_norm": 0.17357373237609863, + "learning_rate": 9.165933312886944e-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": 0.005197142693214118, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0022083621006458998, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.012602647487074136, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.014067314937710762, + "weighted_rkd_loss": 0.006301323743537068 + }, + { + "epoch": 5.538461538461538, + "flow_matching_loss": 0.014604238979518413, + "grad_norm": 0.19311626255512238, + "learning_rate": 9.07466464413155e-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": 0.004733706871047616, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0022366787889041007, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.011704092426225544, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.014604238979518413, + "weighted_rkd_loss": 0.005852046213112772 + }, + { + "epoch": 5.555555555555555, + "flow_matching_loss": 0.015962645784020423, + "grad_norm": 0.1606195569038391, + "learning_rate": 8.983655168740048e-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": 0.004856789810582995, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0018019488430581986, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.011515528429299593, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.015962645784020423, + "weighted_rkd_loss": 0.0057577642146497965 + }, + { + "epoch": 5.572649572649572, + "flow_matching_loss": 0.014079044666141271, + "grad_norm": 0.18407145142555237, + "learning_rate": 8.892908867761718e-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": 0.005349058820866048, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0023212913540191947, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.013019408844411374, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.014079044666141271, + "weighted_rkd_loss": 0.006509704422205687 + }, + { + "epoch": 5.589743589743589, + "flow_matching_loss": 0.014307726826518773, + "grad_norm": 0.1682978719472885, + "learning_rate": 8.802429710733745e-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": 0.005156373279169202, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0019659399753436447, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01227868665009737, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.014307726826518773, + "weighted_rkd_loss": 0.006139343325048685 + }, + { + "epoch": 5.6068376068376065, + "flow_matching_loss": 0.014214032515883445, + "grad_norm": 0.1408603936433792, + "learning_rate": 8.71222165550757e-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": 0.004618929629214108, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0018305400968529283, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.011068399343639613, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.014214032515883445, + "weighted_rkd_loss": 0.0055341996718198064 + }, + { + "epoch": 5.6239316239316235, + "flow_matching_loss": 0.01317118015140295, + "grad_norm": 0.13524463772773743, + "learning_rate": 8.622288648075748e-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": 0.0049630893394351, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0020186342881061137, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.011944812955334783, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.01317118015140295, + "weighted_rkd_loss": 0.005972406477667391 + }, + { + "epoch": 5.641025641025641, + "flow_matching_loss": 0.016691511496901512, + "grad_norm": 0.13098731637001038, + "learning_rate": 8.532634622399391e-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": 0.004113830532878638, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0016622172726783902, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.009889878332614899, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.016691511496901512, + "weighted_rkd_loss": 0.004944939166307449 + }, + { + "epoch": 5.6581196581196584, + "flow_matching_loss": 0.013253772258758545, + "grad_norm": 0.1594141125679016, + "learning_rate": 8.44326350023601e-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": 0.00496185205411166, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0018524963175877928, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.011776200449094176, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.013253772258758545, + "weighted_rkd_loss": 0.005888100224547088 + }, + { + "epoch": 5.6752136752136755, + "flow_matching_loss": 0.01386769926175475, + "grad_norm": 0.18667681515216827, + "learning_rate": 8.354179190968002e-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": 0.0042942218715325, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0017211545608006417, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.010309598175808787, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.01386769926175475, + "weighted_rkd_loss": 0.005154799087904394 + }, + { + "epoch": 5.6923076923076925, + "flow_matching_loss": 0.015427383687347173, + "grad_norm": 0.19144976139068604, + "learning_rate": 8.265385591431661e-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": 0.004844757309183478, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.001974588306620717, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.011664103018119931, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.015427383687347173, + "weighted_rkd_loss": 0.005832051509059966 + }, + { + "epoch": 5.7094017094017095, + "flow_matching_loss": 0.014562000147998333, + "grad_norm": 0.181470587849617, + "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": 0.005005742516368627, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.001948601473122835, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.011960086273029446, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.014562000147998333, + "weighted_rkd_loss": 0.005980043136514723 + }, + { + "epoch": 5.726495726495727, + "flow_matching_loss": 0.01286997003480792, + "grad_norm": 0.2129090130329132, + "learning_rate": 8.088686045146247e-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": 0.00470130869653076, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0017387204454280436, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.011141337687149644, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.01286997003480792, + "weighted_rkd_loss": 0.005570668843574822 + }, + { + "epoch": 5.743589743589744, + "flow_matching_loss": 0.013654280174523591, + "grad_norm": 0.17266348004341125, + "learning_rate": 8.00078782780775e-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": 0.005098359915427864, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.002111944928765297, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.012308664619922638, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.013654280174523591, + "weighted_rkd_loss": 0.006154332309961319 + }, + { + "epoch": 5.760683760683761, + "flow_matching_loss": 0.016567762661725282, + "grad_norm": 0.22100067138671875, + "learning_rate": 7.913195778683979e-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": 0.00517838706728071, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.001889878825750202, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.012246652832254767, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.016567762661725282, + "weighted_rkd_loss": 0.0061233264161273835 + }, + { + "epoch": 5.777777777777778, + "flow_matching_loss": 0.014265890140086413, + "grad_norm": 0.1820782721042633, + "learning_rate": 7.825913729334908e-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": 0.004978736140765249, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.001999659137800336, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.011957131465896964, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.014265890140086413, + "weighted_rkd_loss": 0.005978565732948482 + }, + { + "epoch": 5.794871794871795, + "flow_matching_loss": 0.016519154980778696, + "grad_norm": 0.17449019849300385, + "learning_rate": 7.73894549776013e-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": 0.004733387008309364, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0020300828851759433, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.011496856901794672, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.016519154980778696, + "weighted_rkd_loss": 0.005748428450897336 + }, + { + "epoch": 5.811965811965812, + "flow_matching_loss": 0.01309665311127901, + "grad_norm": 0.1982821375131607, + "learning_rate": 7.65229488823184e-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": 0.00616358332335949, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.002463771589100361, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.014790938142687083, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.01309665311127901, + "weighted_rkd_loss": 0.007395469071343541 + }, + { + "epoch": 5.829059829059829, + "flow_matching_loss": 0.01534111937507987, + "grad_norm": 0.16873861849308014, + "learning_rate": 7.565965691128396e-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": 0.005616517853923142, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0022936845547519625, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01352672027423978, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.01534111937507987, + "weighted_rkd_loss": 0.00676336013711989 + }, + { + "epoch": 5.846153846153846, + "flow_matching_loss": 0.014631201140582561, + "grad_norm": 0.19740276038646698, + "learning_rate": 7.47996168276856e-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": 0.005165145918726921, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0019399866694584488, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.012270278483629226, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.014631201140582561, + "weighted_rkd_loss": 0.006135139241814613 + }, + { + "epoch": 5.863247863247864, + "flow_matching_loss": 0.014703481085598468, + "grad_norm": 0.1830338090658188, + "learning_rate": 7.394286625246258e-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": 0.0049274742603302, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0022761092172004283, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.012131057772785425, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.014703481085598468, + "weighted_rkd_loss": 0.006065528886392712 + }, + { + "epoch": 5.880341880341881, + "flow_matching_loss": 0.013996908906847239, + "grad_norm": 0.12798833847045898, + "learning_rate": 7.308944266266063e-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": 0.00495609906502068, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0020264101680368186, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01193860825151205, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.013996908906847239, + "weighted_rkd_loss": 0.005969304125756025 + }, + { + "epoch": 5.897435897435898, + "flow_matching_loss": 0.015623558219522238, + "grad_norm": 0.17718824744224548, + "learning_rate": 7.223938338979233e-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": 0.004302141885273158, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0017569319461472332, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.010361215798184276, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.015623558219522238, + "weighted_rkd_loss": 0.005180607899092138 + }, + { + "epoch": 5.914529914529915, + "flow_matching_loss": 0.015388756524771452, + "grad_norm": 0.1666608452796936, + "learning_rate": 7.139272561820397e-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": 0.004684569523669779, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0017473775544203818, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01111651649698615, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.015388756524771452, + "weighted_rkd_loss": 0.005558258248493075 + }, + { + "epoch": 5.931623931623932, + "flow_matching_loss": 0.013417072873562575, + "grad_norm": 0.13867004215717316, + "learning_rate": 7.054950638344936e-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": 0.005452374718151987, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0022980000008828937, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.013202749611809849, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.013417072873562575, + "weighted_rkd_loss": 0.0066013748059049245 + }, + { + "epoch": 5.948717948717949, + "flow_matching_loss": 0.012726581562310458, + "grad_norm": 0.14203988015651703, + "learning_rate": 6.970976257066941e-06, + "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": 0.003928991034626961, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0017228475771844387, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.00958082964643836, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.012726581562310458, + "weighted_rkd_loss": 0.00479041482321918 + }, + { + "epoch": 5.965811965811966, + "flow_matching_loss": 0.014724280033260585, + "grad_norm": 0.17691625654697418, + "learning_rate": 6.887353091297896e-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": 0.005078516318462789, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0020477183861657977, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01220475104637444, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.014724280033260585, + "weighted_rkd_loss": 0.00610237552318722 + }, + { + "epoch": 5.982905982905983, + "flow_matching_loss": 0.013537014741450549, + "grad_norm": 0.15324710309505463, + "learning_rate": 6.804084798985965e-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": 0.00474430185277015, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0019807930337265136, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01146939666941762, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.013537014741450549, + "weighted_rkd_loss": 0.00573469833470881 + }, + { + "epoch": 6.0, + "flow_matching_loss": 0.015401409938931465, + "grad_norm": 0.17201311886310577, + "learning_rate": 6.721175022556009e-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": 0.004189198603853583, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0016246947110630573, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01000309200026095, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.015401409938931465, + "weighted_rkd_loss": 0.005001546000130475 + }, + { + "epoch": 6.017094017094017, + "flow_matching_loss": 0.01367185264825821, + "grad_norm": 0.1638801097869873, + "learning_rate": 6.638627388750245e-06, + "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": 0.004297086899168789, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.001833447511307895, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01042762128636241, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.01367185264825821, + "weighted_rkd_loss": 0.005213810643181205 + }, + { + "epoch": 6.034188034188034, + "flow_matching_loss": 0.015124630182981491, + "grad_norm": 0.15200188755989075, + "learning_rate": 6.556445508469583e-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": 0.0053081318968906995, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0021283041802234947, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.012744567822664976, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.015124630182981491, + "weighted_rkd_loss": 0.006372283911332488 + }, + { + "epoch": 6.051282051282051, + "flow_matching_loss": 0.012258633133023977, + "grad_norm": 0.170354962348938, + "learning_rate": 6.474632976615708e-06, + "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": 0.0043724456802010534, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0018084332928992807, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.010553324595093727, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.012258633133023977, + "weighted_rkd_loss": 0.0052766622975468636 + }, + { + "epoch": 6.068376068376068, + "flow_matching_loss": 0.014518317673355341, + "grad_norm": 0.18661901354789734, + "learning_rate": 6.393193371933791e-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": 0.00458554292563349, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0020006953971460463, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.011171781178563833, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.014518317673355341, + "weighted_rkd_loss": 0.005585890589281916 + }, + { + "epoch": 6.085470085470085, + "flow_matching_loss": 0.012813680432736873, + "grad_norm": 0.17715997993946075, + "learning_rate": 6.312130256855977e-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": 0.004458332341164351, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.001910470041912049, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01082713482901454, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.012813680432736873, + "weighted_rkd_loss": 0.00541356741450727 + }, + { + "epoch": 6.102564102564102, + "flow_matching_loss": 0.01346785482019186, + "grad_norm": 0.16243726015090942, + "learning_rate": 6.23144717734552e-06, + "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": 0.004544325242750347, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0018196132208686323, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.010908263735473156, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.01346785482019186, + "weighted_rkd_loss": 0.005454131867736578 + }, + { + "epoch": 6.119658119658119, + "flow_matching_loss": 0.014472407475113868, + "grad_norm": 0.17122097313404083, + "learning_rate": 6.151147662741695e-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": 0.004576569981873035, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.001975224877241999, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.011128364969044924, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.014472407475113868, + "weighted_rkd_loss": 0.005564182484522462 + }, + { + "epoch": 6.136752136752137, + "flow_matching_loss": 0.014526502415537833, + "grad_norm": 0.1407535821199417, + "learning_rate": 6.071235225605415e-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": 0.004797903215512633, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0018253034097142518, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.011421109968796372, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.014526502415537833, + "weighted_rkd_loss": 0.005710554984398186 + }, + { + "epoch": 6.153846153846154, + "flow_matching_loss": 0.013544728001579642, + "grad_norm": 0.13234646618366241, + "learning_rate": 5.991713361565546e-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": 0.0049188114702701565, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0021697683492675423, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.012007391452789307, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.013544728001579642, + "weighted_rkd_loss": 0.0060036957263946535 + }, + { + "epoch": 6.170940170940171, + "flow_matching_loss": 0.013128768280148506, + "grad_norm": 0.17086155712604523, + "learning_rate": 5.912585549166051e-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": 0.004967864602804184, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0018786639207974075, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.011814393242821098, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.013128768280148506, + "weighted_rkd_loss": 0.005907196621410549 + }, + { + "epoch": 6.188034188034188, + "flow_matching_loss": 0.013461261987686157, + "grad_norm": 0.17070406675338745, + "learning_rate": 5.833855249713777e-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": 0.004719796217978, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.001841046300251037, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.011280638724565506, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.013461261987686157, + "weighted_rkd_loss": 0.005640319362282753 + }, + { + "epoch": 6.205128205128205, + "flow_matching_loss": 0.014641128573566675, + "grad_norm": 0.19366569817066193, + "learning_rate": 5.7555259071270645e-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": 0.004503894480876625, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0017051879549399018, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.010712976986542343, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.014641128573566675, + "weighted_rkd_loss": 0.005356488493271172 + }, + { + "epoch": 6.222222222222222, + "flow_matching_loss": 0.012268691649660468, + "grad_norm": 0.17229615151882172, + "learning_rate": 5.677600947785117e-06, + "loss": 0.0179, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.004729417897760868, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0018767193774692713, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.011335555091500282, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.012268691649660468, + "weighted_rkd_loss": 0.005667777545750141 + }, + { + "epoch": 6.239316239316239, + "flow_matching_loss": 0.013378062937408685, + "grad_norm": 0.18267109990119934, + "learning_rate": 5.600083780378099e-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": 0.004501617699861527, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.001768733304925263, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.010771968681365252, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.013378062937408685, + "weighted_rkd_loss": 0.005385984340682626 + }, + { + "epoch": 6.256410256410256, + "flow_matching_loss": 0.01337980325333774, + "grad_norm": 0.2579333782196045, + "learning_rate": 5.522977795758038e-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": 0.004270854382775724, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.001658515038434416, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.010200223792344332, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.01337980325333774, + "weighted_rkd_loss": 0.005100111896172166 + }, + { + "epoch": 6.273504273504273, + "flow_matching_loss": 0.014495591539889574, + "grad_norm": 0.19292736053466797, + "learning_rate": 5.4462863667904805e-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": 0.005087881651706994, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0017388155218213796, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.011914578825235366, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.014495591539889574, + "weighted_rkd_loss": 0.005957289412617683 + }, + { + "epoch": 6.2905982905982905, + "flow_matching_loss": 0.014014208316802978, + "grad_norm": 0.2239702343940735, + "learning_rate": 5.370012848206983e-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": 0.004500076500698924, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0018213755334727465, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.010821528499945998, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.014014208316802978, + "weighted_rkd_loss": 0.005410764249972999 + }, + { + "epoch": 6.3076923076923075, + "flow_matching_loss": 0.0140917822252959, + "grad_norm": 0.17948734760284424, + "learning_rate": 5.29416057645834e-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": 0.0042427378240972756, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0016766243032179772, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.010162099963054061, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.0140917822252959, + "weighted_rkd_loss": 0.0050810499815270305 + }, + { + "epoch": 6.3247863247863245, + "flow_matching_loss": 0.014403108693659305, + "grad_norm": 0.15205460786819458, + "learning_rate": 5.218732869568635e-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": 0.0042243374977260825, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0018360894406214356, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.010284764459356666, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.014403108693659305, + "weighted_rkd_loss": 0.005142382229678333 + }, + { + "epoch": 6.3418803418803416, + "flow_matching_loss": 0.012675300985574723, + "grad_norm": 0.1909080594778061, + "learning_rate": 5.14373302699013e-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": 0.005056148441508412, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.00219099071691744, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.012303287675604225, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.012675300985574723, + "weighted_rkd_loss": 0.0061516438378021124 + }, + { + "epoch": 6.358974358974359, + "flow_matching_loss": 0.014092040248215199, + "grad_norm": 0.17801937460899353, + "learning_rate": 5.069164329458894e-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": 0.004562873602844774, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0017242739209905266, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.010850021056830883, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.014092040248215199, + "weighted_rkd_loss": 0.005425010528415442 + }, + { + "epoch": 6.3760683760683765, + "flow_matching_loss": 0.014610704127699136, + "grad_norm": 0.16385993361473083, + "learning_rate": 4.995030038851342e-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": 0.004508930863812566, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0017317418241873384, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.010749603435397149, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.014610704127699136, + "weighted_rkd_loss": 0.005374801717698574 + }, + { + "epoch": 6.3931623931623935, + "flow_matching_loss": 0.011634174268692732, + "grad_norm": 0.15849269926548004, + "learning_rate": 4.9213333980415e-06, + "loss": 0.0166, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.004152471316047013, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.001584779960103333, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.009889722522348165, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.011634174268692732, + "weighted_rkd_loss": 0.004944861261174082 + }, + { + "epoch": 6.410256410256411, + "flow_matching_loss": 0.013286026380956172, + "grad_norm": 0.1140068992972374, + "learning_rate": 4.848077630759175e-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": 0.00479142447002232, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0021043072978500278, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.011687156278640032, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.013286026380956172, + "weighted_rkd_loss": 0.005843578139320016 + }, + { + "epoch": 6.427350427350428, + "flow_matching_loss": 0.014323109947144985, + "grad_norm": 0.17702610790729523, + "learning_rate": 4.775265941448951e-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": 0.005099014192819595, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0019039258710108697, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.012101954314857721, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.014323109947144985, + "weighted_rkd_loss": 0.006050977157428861 + }, + { + "epoch": 6.444444444444445, + "flow_matching_loss": 0.013860237039625644, + "grad_norm": 0.1613358110189438, + "learning_rate": 4.702901515129989e-06, + "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": 0.004091036505997181, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.001842647383455187, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.010024720383808017, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.013860237039625644, + "weighted_rkd_loss": 0.005012360191904009 + }, + { + "epoch": 6.461538461538462, + "flow_matching_loss": 0.013166516833007336, + "grad_norm": 0.12258017063140869, + "learning_rate": 4.630987517256734e-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": 0.005356473685242236, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.00200028627878055, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.012713233754038811, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.013166516833007336, + "weighted_rkd_loss": 0.006356616877019406 + }, + { + "epoch": 6.478632478632479, + "flow_matching_loss": 0.012790456414222717, + "grad_norm": 0.1847393661737442, + "learning_rate": 4.559527093580412e-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": 0.004619638319127262, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0019545887014828624, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.011193865444511175, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.012790456414222717, + "weighted_rkd_loss": 0.005596932722255588 + }, + { + "epoch": 6.495726495726496, + "flow_matching_loss": 0.012947164382785559, + "grad_norm": 0.16731011867523193, + "learning_rate": 4.488523370011456e-06, + "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": 0.004310220014303923, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0016456021461635829, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.010266042361035944, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.012947164382785559, + "weighted_rkd_loss": 0.005133021180517972 + }, + { + "epoch": 6.512820512820513, + "flow_matching_loss": 0.012447025114670395, + "grad_norm": 0.13510732352733612, + "learning_rate": 4.417979452482766e-06, + "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": 0.004228096292354167, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0016668969125021248, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.010123089654371142, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.012447025114670395, + "weighted_rkd_loss": 0.005061544827185571 + }, + { + "epoch": 6.52991452991453, + "flow_matching_loss": 0.01330544799566269, + "grad_norm": 0.14951033890247345, + "learning_rate": 4.347898426813811e-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": 0.00396587192080915, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0014627483498770744, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.009394492162391544, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.01330544799566269, + "weighted_rkd_loss": 0.004697246081195772 + }, + { + "epoch": 6.547008547008547, + "flow_matching_loss": 0.01360499719157815, + "grad_norm": 0.16325245797634125, + "learning_rate": 4.278283358575694e-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": 0.004816550016403198, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0018093404709361494, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.011442440561950207, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.01360499719157815, + "weighted_rkd_loss": 0.005721220280975103 + }, + { + "epoch": 6.564102564102564, + "flow_matching_loss": 0.01414185781031847, + "grad_norm": 0.2390795648097992, + "learning_rate": 4.2091372929570145e-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": 0.004850478307344019, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.002294772188179195, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.011995728733018041, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.01414185781031847, + "weighted_rkd_loss": 0.0059978643665090205 + }, + { + "epoch": 6.581196581196581, + "flow_matching_loss": 0.016646978538483383, + "grad_norm": 0.1479363888502121, + "learning_rate": 4.1404632546306934e-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": 0.0038240491645410656, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0016426807967945933, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.009290779009461403, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.016646978538483383, + "weighted_rkd_loss": 0.0046453895047307014 + }, + { + "epoch": 6.598290598290598, + "flow_matching_loss": 0.013259697053581477, + "grad_norm": 0.1548348218202591, + "learning_rate": 4.0722642476216286e-06, + "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": 0.004015104891732335, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0015654072863981127, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.00959561700001359, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.013259697053581477, + "weighted_rkd_loss": 0.004797808500006795 + }, + { + "epoch": 6.615384615384615, + "flow_matching_loss": 0.013232738617807627, + "grad_norm": 0.17228160798549652, + "learning_rate": 4.00454325517533e-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": 0.0046642918838188056, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0017990636872127653, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.011127647291868926, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.013232738617807627, + "weighted_rkd_loss": 0.005563823645934463 + }, + { + "epoch": 6.632478632478632, + "flow_matching_loss": 0.013004853390157223, + "grad_norm": 0.17855753004550934, + "learning_rate": 3.937303239627393e-06, + "loss": 0.017, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.003314048214815557, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0014266120444517583, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.008054708503186703, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.013004853390157223, + "weighted_rkd_loss": 0.0040273542515933515 + }, + { + "epoch": 6.64957264957265, + "flow_matching_loss": 0.014002307411283254, + "grad_norm": 0.24588173627853394, + "learning_rate": 3.870547142273918e-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": 0.00451577752828598, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0020319839182775467, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.011063539003953338, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.014002307411283254, + "weighted_rkd_loss": 0.005531769501976669 + }, + { + "epoch": 6.666666666666667, + "flow_matching_loss": 0.013700271677225827, + "grad_norm": 0.1725107878446579, + "learning_rate": 3.8042778832428736e-06, + "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": 0.0036921764723956583, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0014130157709587366, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.008797368640080095, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.013700271677225827, + "weighted_rkd_loss": 0.004398684320040047 + }, + { + "epoch": 6.683760683760684, + "flow_matching_loss": 0.01752990819513798, + "grad_norm": 0.1484254151582718, + "learning_rate": 3.7384983613663195e-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": 0.004090518108569086, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0015098612115252764, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.00969089725986123, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.01752990819513798, + "weighted_rkd_loss": 0.004845448629930615 + }, + { + "epoch": 6.700854700854701, + "flow_matching_loss": 0.015219774376600981, + "grad_norm": 0.21970634162425995, + "learning_rate": 3.6732114540536503e-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": 0.004155363631434738, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0016006131307221949, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.009911340568214655, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.015219774376600981, + "weighted_rkd_loss": 0.004955670284107328 + }, + { + "epoch": 6.717948717948718, + "flow_matching_loss": 0.014597513433545827, + "grad_norm": 0.1565510630607605, + "learning_rate": 3.608420017165683e-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": 0.004712637630291283, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0017746063880622388, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.011199881695210934, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.014597513433545827, + "weighted_rkd_loss": 0.005599940847605467 + }, + { + "epoch": 6.735042735042735, + "flow_matching_loss": 0.015428585838526488, + "grad_norm": 0.21946097910404205, + "learning_rate": 3.5441268848897664e-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": 0.004173310683108866, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0015550711716059596, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.009901692438870669, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.015428585838526488, + "weighted_rkd_loss": 0.0049508462194353346 + }, + { + "epoch": 6.752136752136752, + "flow_matching_loss": 0.012527666334062815, + "grad_norm": 0.13608911633491516, + "learning_rate": 3.480334869615792e-06, + "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": 0.004187409998849034, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.001549099455587566, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.009923919336870313, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.012527666334062815, + "weighted_rkd_loss": 0.004961959668435156 + }, + { + "epoch": 6.769230769230769, + "flow_matching_loss": 0.01083629266358912, + "grad_norm": 0.18144173920154572, + "learning_rate": 3.417046761813155e-06, + "loss": 0.0163, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.004622485325671733, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0017338956182356923, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.010978866182267667, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.01083629266358912, + "weighted_rkd_loss": 0.005489433091133833 + }, + { + "epoch": 6.786324786324786, + "flow_matching_loss": 0.014319020509719848, + "grad_norm": 0.21980780363082886, + "learning_rate": 3.3542653299087263e-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": 0.004227689397521317, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0015997855283785612, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.010055164387449623, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.014319020509719848, + "weighted_rkd_loss": 0.005027582193724811 + }, + { + "epoch": 6.803418803418803, + "flow_matching_loss": 0.010222503775730729, + "grad_norm": 0.1556553691625595, + "learning_rate": 3.2919933201657117e-06, + "loss": 0.0148, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.0037521095015108584, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0016987596172839403, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.009202978760004043, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.010222503775730729, + "weighted_rkd_loss": 0.0046014893800020214 + }, + { + "epoch": 6.82051282051282, + "flow_matching_loss": 0.014451631903648376, + "grad_norm": 0.19691146910190582, + "learning_rate": 3.2302334565635584e-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": 0.004138553678058088, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0016731622163206339, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.009950269386172295, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.014451631903648376, + "weighted_rkd_loss": 0.0049751346930861475 + }, + { + "epoch": 6.837606837606837, + "flow_matching_loss": 0.011851238133385778, + "grad_norm": 0.20504842698574066, + "learning_rate": 3.1689884406787674e-06, + "loss": 0.0163, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.0037106047151610256, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.001522247534012422, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.008943456877022981, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.011851238133385778, + "weighted_rkd_loss": 0.0044717284385114905 + }, + { + "epoch": 6.854700854700854, + "flow_matching_loss": 0.014245664700865746, + "grad_norm": 0.14268416166305542, + "learning_rate": 3.1082609515667424e-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": 0.003939698520116508, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0016180734150111676, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.009497470408678054, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.014245664700865746, + "weighted_rkd_loss": 0.004748735204339027 + }, + { + "epoch": 6.871794871794872, + "flow_matching_loss": 0.014305701572448015, + "grad_norm": 0.22521346807479858, + "learning_rate": 3.048053645644594e-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": 0.005096148257143796, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0019075013406109065, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01209979779087007, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.014305701572448015, + "weighted_rkd_loss": 0.006049898895435035 + }, + { + "epoch": 6.888888888888889, + "flow_matching_loss": 0.013212948385626078, + "grad_norm": 0.148503378033638, + "learning_rate": 2.988369156574916e-06, + "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": 0.003819907363504171, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0015058944351039826, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.0091457090806216, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.013212948385626078, + "weighted_rkd_loss": 0.0045728545403108 + }, + { + "epoch": 6.905982905982906, + "flow_matching_loss": 0.01349337175488472, + "grad_norm": 0.13594625890254974, + "learning_rate": 2.9292100951506225e-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": 0.004378201114013791, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0016772790113463998, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.010433681216090918, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.01349337175488472, + "weighted_rkd_loss": 0.005216840608045459 + }, + { + "epoch": 6.923076923076923, + "flow_matching_loss": 0.013800859311595559, + "grad_norm": 0.17430321872234344, + "learning_rate": 2.8705790491807033e-06, + "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": 0.004294183477759362, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0016260517295449972, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01021441868506372, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.013800859311595559, + "weighted_rkd_loss": 0.00510720934253186 + }, + { + "epoch": 6.94017094017094, + "flow_matching_loss": 0.011465824535116553, + "grad_norm": 0.18166108429431915, + "learning_rate": 2.8124785833770433e-06, + "loss": 0.0165, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.004202205874025822, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0016021716815885156, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.010006583435460925, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.011465824535116553, + "weighted_rkd_loss": 0.005003291717730463 + }, + { + "epoch": 6.957264957264957, + "flow_matching_loss": 0.01547788204625249, + "grad_norm": 0.1667901873588562, + "learning_rate": 2.754911239242241e-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": 0.0038020822452381254, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0018294627370778471, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.009433627175167204, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.01547788204625249, + "weighted_rkd_loss": 0.004716813587583602 + }, + { + "epoch": 6.9743589743589745, + "flow_matching_loss": 0.012793379882350565, + "grad_norm": 0.15985463559627533, + "learning_rate": 2.6978795349584204e-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": 0.004487853916361928, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0018074786174111068, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.010783186368644237, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.012793379882350565, + "weighted_rkd_loss": 0.005391593184322118 + }, + { + "epoch": 6.9914529914529915, + "flow_matching_loss": 0.013774941116571427, + "grad_norm": 0.16853109002113342, + "learning_rate": 2.641385965277088e-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": 0.004574715811759234, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0018272122426424176, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01097664381377399, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.013774941116571427, + "weighted_rkd_loss": 0.005488321906886995 + }, + { + "epoch": 7.0085470085470085, + "flow_matching_loss": 0.01377521250396967, + "grad_norm": 0.16738949716091156, + "learning_rate": 2.5854330014099923e-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": 0.004262465052306652, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0014018778572790324, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.009926808020099997, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.01377521250396967, + "weighted_rkd_loss": 0.004963404010049998 + }, + { + "epoch": 7.0256410256410255, + "flow_matching_loss": 0.013304527848958969, + "grad_norm": 0.1807527244091034, + "learning_rate": 2.5300230909210298e-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": 0.004257907392457128, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0017975858820136637, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.010313400626182556, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.013304527848958969, + "weighted_rkd_loss": 0.005156700313091278 + }, + { + "epoch": 7.042735042735043, + "flow_matching_loss": 0.012193146161735058, + "grad_norm": 0.2061319798231125, + "learning_rate": 2.4751586576191983e-06, + "loss": 0.0171, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.0041587944375351075, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.001565186760853976, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.009882775600999594, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.012193146161735058, + "weighted_rkd_loss": 0.004941387800499797 + }, + { + "epoch": 7.05982905982906, + "flow_matching_loss": 0.01150200036354363, + "grad_norm": 0.12688463926315308, + "learning_rate": 2.420842101452538e-06, + "loss": 0.0163, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.003888703393749893, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0018680165347177535, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.009645423293113709, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.01150200036354363, + "weighted_rkd_loss": 0.004822711646556854 + }, + { + "epoch": 7.076923076923077, + "flow_matching_loss": 0.011715107830241323, + "grad_norm": 0.17974625527858734, + "learning_rate": 2.3670757984031816e-06, + "loss": 0.0162, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.003783612954430282, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0014903668721672147, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.009057592926546931, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.011715107830241323, + "weighted_rkd_loss": 0.0045287964632734655 + }, + { + "epoch": 7.094017094017094, + "flow_matching_loss": 0.011287252325564623, + "grad_norm": 0.1417006403207779, + "learning_rate": 2.313862100383404e-06, + "loss": 0.0154, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.00343161195050925, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0013535286823753267, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.008216752624139189, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.011287252325564623, + "weighted_rkd_loss": 0.0041083763120695945 + }, + { + "epoch": 7.111111111111111, + "flow_matching_loss": 0.012390893045812845, + "grad_norm": 0.1228102594614029, + "learning_rate": 2.2612033351327406e-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": 0.0042154436698183416, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0019080315832979977, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.010338918957859277, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.012390893045812845, + "weighted_rkd_loss": 0.0051694594789296385 + }, + { + "epoch": 7.128205128205128, + "flow_matching_loss": 0.011362318880856037, + "grad_norm": 0.1354866623878479, + "learning_rate": 2.2091018061161788e-06, + "loss": 0.016, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.0038231971440836785, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0017028606845997274, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.009349254984408617, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.011362318880856037, + "weighted_rkd_loss": 0.004674627492204308 + }, + { + "epoch": 7.145299145299146, + "flow_matching_loss": 0.014426226122304797, + "grad_norm": 0.14201338589191437, + "learning_rate": 2.15755979242338e-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": 0.004279016936197877, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.001714422053191811, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.010272455867379904, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.014426226122304797, + "weighted_rkd_loss": 0.005136227933689952 + }, + { + "epoch": 7.162393162393163, + "flow_matching_loss": 0.013876243494451046, + "grad_norm": 0.15545055270195007, + "learning_rate": 2.106579548669e-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": 0.003654197743162513, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0012436681019607932, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.008552063629031182, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.013876243494451046, + "weighted_rkd_loss": 0.004276031814515591 + }, + { + "epoch": 7.17948717948718, + "flow_matching_loss": 0.012172015756368637, + "grad_norm": 0.18450547754764557, + "learning_rate": 2.0561633048940477e-06, + "loss": 0.0172, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.004115495644509792, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0017840529326349497, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.010015044175088406, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.012172015756368637, + "weighted_rkd_loss": 0.005007522087544203 + }, + { + "epoch": 7.196581196581197, + "flow_matching_loss": 0.01326232748106122, + "grad_norm": 0.1385352909564972, + "learning_rate": 2.006313266468355e-06, + "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": 0.004702828754670918, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0018367817858234048, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.011242439271882176, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.01326232748106122, + "weighted_rkd_loss": 0.005621219635941088 + }, + { + "epoch": 7.213675213675214, + "flow_matching_loss": 0.01336014000698924, + "grad_norm": 0.16189557313919067, + "learning_rate": 1.957031613994097e-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": 0.003861038712784648, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0015294211858417838, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.009251498710364103, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.01336014000698924, + "weighted_rkd_loss": 0.0046257493551820515 + }, + { + "epoch": 7.230769230769231, + "flow_matching_loss": 0.013220749609172345, + "grad_norm": 0.18385490775108337, + "learning_rate": 1.9083205032103995e-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": 0.0037725105416029693, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0013661614037118852, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.008911182498559355, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.013220749609172345, + "weighted_rkd_loss": 0.004455591249279678 + }, + { + "epoch": 7.247863247863248, + "flow_matching_loss": 0.01324486918747425, + "grad_norm": 0.18078869581222534, + "learning_rate": 1.8601820648990536e-06, + "loss": 0.0179, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.0039695461047813295, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0013943981553893535, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.009333490440621973, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.01324486918747425, + "weighted_rkd_loss": 0.004666745220310986 + }, + { + "epoch": 7.264957264957265, + "flow_matching_loss": 0.01174166426062584, + "grad_norm": 0.14929896593093872, + "learning_rate": 1.8126184047912936e-06, + "loss": 0.0164, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.0038963052444159984, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0016135171288624405, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.009406127594411373, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.01174166426062584, + "weighted_rkd_loss": 0.004703063797205686 + }, + { + "epoch": 7.282051282051282, + "flow_matching_loss": 0.01100527481175959, + "grad_norm": 0.2219940721988678, + "learning_rate": 1.7656316034757024e-06, + "loss": 0.0167, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.004765593982301653, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0019187520316336303, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.011449940036982298, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.01100527481175959, + "weighted_rkd_loss": 0.005724970018491149 + }, + { + "epoch": 7.299145299145299, + "flow_matching_loss": 0.013349955342710019, + "grad_norm": 0.12957392632961273, + "learning_rate": 1.7192237163071816e-06, + "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": 0.003737821336835623, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.001505720429122448, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.008981363102793693, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.013349955342710019, + "weighted_rkd_loss": 0.004490681551396846 + }, + { + "epoch": 7.316239316239316, + "flow_matching_loss": 0.012917829304933548, + "grad_norm": 0.13917148113250732, + "learning_rate": 1.6733967733170573e-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": 0.0037979158572852612, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.002031996531877667, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.009627828234806656, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.012917829304933548, + "weighted_rkd_loss": 0.004813914117403328 + }, + { + "epoch": 7.333333333333333, + "flow_matching_loss": 0.011703525995835662, + "grad_norm": 0.19382449984550476, + "learning_rate": 1.628152779124279e-06, + "loss": 0.0166, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.004080761969089508, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.001575100829359144, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.009736624732613563, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.011703525995835662, + "weighted_rkd_loss": 0.004868312366306781 + }, + { + "epoch": 7.35042735042735, + "flow_matching_loss": 0.01296322252601385, + "grad_norm": 0.14509230852127075, + "learning_rate": 1.583493712847715e-06, + "loss": 0.0174, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.0037901628063991668, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0013702258991543203, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.008950551599264145, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.01296322252601385, + "weighted_rkd_loss": 0.004475275799632072 + }, + { + "epoch": 7.367521367521368, + "flow_matching_loss": 0.014123854786157608, + "grad_norm": 0.1688782125711441, + "learning_rate": 1.5394215280196084e-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": 0.0036876884289085863, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0015890191541984677, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.008964395988732577, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.014123854786157608, + "weighted_rkd_loss": 0.004482197994366288 + }, + { + "epoch": 7.384615384615385, + "flow_matching_loss": 0.013744353968650102, + "grad_norm": 0.23357872664928436, + "learning_rate": 1.4959381525000904e-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": 0.0046668478287756445, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0017815333674661815, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01111522908322513, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.013744353968650102, + "weighted_rkd_loss": 0.005557614541612565 + }, + { + "epoch": 7.401709401709402, + "flow_matching_loss": 0.013115393370389939, + "grad_norm": 0.13692398369312286, + "learning_rate": 1.453045488392879e-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": 0.004141918360255658, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.001547129702521488, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.009830966452136635, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.013115393370389939, + "weighted_rkd_loss": 0.004915483226068318 + }, + { + "epoch": 7.418803418803419, + "flow_matching_loss": 0.012881484348326921, + "grad_norm": 0.20485293865203857, + "learning_rate": 1.4107454119620484e-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": 0.003988063824363053, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0015699075884185731, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.009546035202220082, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.012881484348326921, + "weighted_rkd_loss": 0.004773017601110041 + }, + { + "epoch": 7.435897435897436, + "flow_matching_loss": 0.01345738205127418, + "grad_norm": 0.2101745456457138, + "learning_rate": 1.3690397735499778e-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": 0.003833774570375681, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0017207548720762134, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.009388304036110639, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.01345738205127418, + "weighted_rkd_loss": 0.0046941520180553194 + }, + { + "epoch": 7.452991452991453, + "flow_matching_loss": 0.014391318708658219, + "grad_norm": 0.17784802615642548, + "learning_rate": 1.327930397496403e-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": 0.004032033658586442, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.001528675900772214, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.009592743217945099, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.014391318708658219, + "weighted_rkd_loss": 0.004796371608972549 + }, + { + "epoch": 7.47008547008547, + "flow_matching_loss": 0.011513168923556805, + "grad_norm": 0.18674953281879425, + "learning_rate": 1.287419082058604e-06, + "loss": 0.0159, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.0035867713391780852, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0016056925931479781, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.008779235230758787, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.011513168923556805, + "weighted_rkd_loss": 0.0043896176153793934 + }, + { + "epoch": 7.487179487179487, + "flow_matching_loss": 0.012771233264356852, + "grad_norm": 0.16635802388191223, + "learning_rate": 1.2475075993327589e-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": 0.004345070035196841, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.00172021365724504, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.010410353634506464, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.012771233264356852, + "weighted_rkd_loss": 0.005205176817253232 + }, + { + "epoch": 7.504273504273504, + "flow_matching_loss": 0.013028046255931259, + "grad_norm": 0.11585721373558044, + "learning_rate": 1.208197695176413e-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": 0.004318495141342282, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0016846387938130648, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.010321629187092184, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.013028046255931259, + "weighted_rkd_loss": 0.005160814593546092 + }, + { + "epoch": 7.521367521367521, + "flow_matching_loss": 0.011222801310941577, + "grad_norm": 0.14337489008903503, + "learning_rate": 1.1694910891321259e-06, + "loss": 0.016, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.003986520576290786, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0016476838150992989, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.009620724944397807, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.011222801310941577, + "weighted_rkd_loss": 0.0048103624721989036 + }, + { + "epoch": 7.538461538461538, + "flow_matching_loss": 0.014170465990900993, + "grad_norm": 0.1666136533021927, + "learning_rate": 1.1313894743522363e-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": 0.004018307896330953, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0017779929097741843, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.00981460870243609, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.014170465990900993, + "weighted_rkd_loss": 0.004907304351218045 + }, + { + "epoch": 7.555555555555555, + "flow_matching_loss": 0.012973797554150224, + "grad_norm": 0.1495077759027481, + "learning_rate": 1.0938945175248016e-06, + "loss": 0.0172, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.00348243482876569, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.001476709794951603, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.008441579481586813, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.012973797554150224, + "weighted_rkd_loss": 0.004220789740793407 + }, + { + "epoch": 7.572649572649572, + "flow_matching_loss": 0.013660789560526609, + "grad_norm": 0.1573425829410553, + "learning_rate": 1.0570078588007086e-06, + "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": 0.0042850013822317125, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0018650735961273312, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.010435076337307692, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.013660789560526609, + "weighted_rkd_loss": 0.005217538168653846 + }, + { + "epoch": 7.589743589743589, + "flow_matching_loss": 0.010968834720551967, + "grad_norm": 0.17911341786384583, + "learning_rate": 1.0207311117219009e-06, + "loss": 0.0156, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.0039019836112856865, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0014731241972185671, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.009277091501280666, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.010968834720551967, + "weighted_rkd_loss": 0.004638545750640333 + }, + { + "epoch": 7.6068376068376065, + "flow_matching_loss": 0.012092063948512078, + "grad_norm": 0.21830224990844727, + "learning_rate": 9.850658631508198e-07, + "loss": 0.0172, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.004266565758734941, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0017167957383207976, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.010249927174299955, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.012092063948512078, + "weighted_rkd_loss": 0.005124963587149978 + }, + { + "epoch": 7.6239316239316235, + "flow_matching_loss": 0.013988563138991595, + "grad_norm": 0.14057599008083344, + "learning_rate": 9.500136732009751e-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": 0.0038496993482112886, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0015126889105886222, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.009212087607011199, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.013988563138991595, + "weighted_rkd_loss": 0.0046060438035055995 + }, + { + "epoch": 7.641025641025641, + "flow_matching_loss": 0.010955549450591207, + "grad_norm": 0.18209223449230194, + "learning_rate": 9.155760751687054e-07, + "loss": 0.0157, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.003972943243570626, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0015911909751594067, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.009537077462300658, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.010955549450591207, + "weighted_rkd_loss": 0.004768538731150329 + }, + { + "epoch": 7.6581196581196584, + "flow_matching_loss": 0.012481959071010351, + "grad_norm": 0.21544641256332397, + "learning_rate": 8.817545754661155e-07, + "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": 0.004292281693778932, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0016132907243445516, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01019785408861935, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.012481959071010351, + "weighted_rkd_loss": 0.005098927044309675 + }, + { + "epoch": 7.6752136752136755, + "flow_matching_loss": 0.014317840617150069, + "grad_norm": 0.13255079090595245, + "learning_rate": 8.485506535551668e-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": 0.0034783784998580814, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0013597151497378944, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.008316472126170992, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.014317840617150069, + "weighted_rkd_loss": 0.004158236063085496 + }, + { + "epoch": 7.6923076923076925, + "flow_matching_loss": 0.010459744790568947, + "grad_norm": 0.17515155673027039, + "learning_rate": 8.159657618829685e-07, + "loss": 0.0152, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.004000649321824312, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0014889903599396349, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.009490288980305194, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.010459744790568947, + "weighted_rkd_loss": 0.004745144490152597 + }, + { + "epoch": 7.7094017094017095, + "flow_matching_loss": 0.01422869712114334, + "grad_norm": 0.17458616197109222, + "learning_rate": 7.840013258182471e-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": 0.003670058771967888, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.001456214237259701, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.008796331845223904, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.01422869712114334, + "weighted_rkd_loss": 0.004398165922611952 + }, + { + "epoch": 7.726495726495727, + "flow_matching_loss": 0.013734146859496832, + "grad_norm": 0.14348195493221283, + "learning_rate": 7.526587435889809e-07, + "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": 0.003523202263750136, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.001518395874882117, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.008564800256863236, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.013734146859496832, + "weighted_rkd_loss": 0.004282400128431618 + }, + { + "epoch": 7.743589743589744, + "flow_matching_loss": 0.01234424621798098, + "grad_norm": 0.1296064853668213, + "learning_rate": 7.219393862212593e-07, + "loss": 0.0167, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.003714152844622731, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0013201057794503868, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.008748411480337382, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.01234424621798098, + "weighted_rkd_loss": 0.004374205740168691 + }, + { + "epoch": 7.760683760683761, + "flow_matching_loss": 0.012772375950589776, + "grad_norm": 0.1534041166305542, + "learning_rate": 6.918445974792819e-07, + "loss": 0.0165, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.0030991335166618227, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0011701180832460524, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.007368385139852762, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.012772375950589776, + "weighted_rkd_loss": 0.003684192569926381 + }, + { + "epoch": 7.777777777777778, + "flow_matching_loss": 0.012171571562066674, + "grad_norm": 0.10533745586872101, + "learning_rate": 6.6237569380661e-07, + "loss": 0.0166, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.003707205457612872, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0014780213998164982, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.008892432367429138, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.012171571562066674, + "weighted_rkd_loss": 0.004446216183714569 + }, + { + "epoch": 7.794871794871795, + "flow_matching_loss": 0.012081097718328238, + "grad_norm": 0.11634702980518341, + "learning_rate": 6.335339642685495e-07, + "loss": 0.0163, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.0034490347607061267, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.001452084380434826, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.008350153779610991, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.012081097718328238, + "weighted_rkd_loss": 0.0041750768898054956 + }, + { + "epoch": 7.811965811965812, + "flow_matching_loss": 0.013012279709801077, + "grad_norm": 0.18919211626052856, + "learning_rate": 6.05320670495787e-07, + "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": 0.004376333625987172, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0016027019592002034, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.010355369048193097, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.013012279709801077, + "weighted_rkd_loss": 0.005177684524096548 + }, + { + "epoch": 7.829059829059829, + "flow_matching_loss": 0.011991818202659488, + "grad_norm": 0.17841693758964539, + "learning_rate": 5.777370466291954e-07, + "loss": 0.0166, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.0039478613995015625, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0013961193850263952, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.00929184230044484, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.011991818202659488, + "weighted_rkd_loss": 0.00464592115022242 + }, + { + "epoch": 7.846153846153846, + "flow_matching_loss": 0.010694775450974702, + "grad_norm": 0.15433743596076965, + "learning_rate": 5.507842992658336e-07, + "loss": 0.0148, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.0033815303584560754, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0013989461411256343, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.008162006922066211, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.010694775450974702, + "weighted_rkd_loss": 0.0040810034610331055 + }, + { + "epoch": 7.863247863247864, + "flow_matching_loss": 0.014666303992271423, + "grad_norm": 0.1808875948190689, + "learning_rate": 5.24463607406192e-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": 0.00379934289958328, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0016835816204547882, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.009282267419621348, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.014666303992271423, + "weighted_rkd_loss": 0.004641133709810674 + }, + { + "epoch": 7.880341880341881, + "flow_matching_loss": 0.011279391311109065, + "grad_norm": 0.17627708613872528, + "learning_rate": 4.987761224025944e-07, + "loss": 0.0163, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.004237288725562394, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0014775051502510905, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.009952082624658942, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.011279391311109065, + "weighted_rkd_loss": 0.004976041312329471 + }, + { + "epoch": 7.897435897435898, + "flow_matching_loss": 0.015158367482945323, + "grad_norm": 0.14843781292438507, + "learning_rate": 4.737229679088556e-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": 0.004200689448043704, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.00160459796898067, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01000597677193582, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.015158367482945323, + "weighted_rkd_loss": 0.00500298838596791 + }, + { + "epoch": 7.914529914529915, + "flow_matching_loss": 0.013612598646432161, + "grad_norm": 0.1926652193069458, + "learning_rate": 4.493052398311115e-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": 0.0039387172088027, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0016255067079328001, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.009502941137179732, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.013612598646432161, + "weighted_rkd_loss": 0.004751470568589866 + }, + { + "epoch": 7.931623931623932, + "flow_matching_loss": 0.0120464526116848, + "grad_norm": 0.16931195557117462, + "learning_rate": 4.2552400627989043e-07, + "loss": 0.0168, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.004019210627302527, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0015675419068429618, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.009605963109061122, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.0120464526116848, + "weighted_rkd_loss": 0.004802981554530561 + }, + { + "epoch": 7.948717948717949, + "flow_matching_loss": 0.012973124394193292, + "grad_norm": 0.1812753826379776, + "learning_rate": 4.023803075233923e-07, + "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": 0.004355783481150866, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0017748346959706395, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.010486401710659265, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.012973124394193292, + "weighted_rkd_loss": 0.005243200855329633 + }, + { + "epoch": 7.965811965811966, + "flow_matching_loss": 0.013465069141238928, + "grad_norm": 0.16017380356788635, + "learning_rate": 3.7987515594196997e-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": 0.0038325405679643155, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.001507552177645266, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.009172633336856962, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.013465069141238928, + "weighted_rkd_loss": 0.004586316668428481 + }, + { + "epoch": 7.982905982905983, + "flow_matching_loss": 0.012850679084658622, + "grad_norm": 0.14726558327674866, + "learning_rate": 3.580095359838648e-07, + "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": 0.004245123150758446, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0017716314876452088, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.010261877812445164, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.012850679084658622, + "weighted_rkd_loss": 0.005130938906222582 + }, + { + "epoch": 8.0, + "flow_matching_loss": 0.013819062430411578, + "grad_norm": 0.14340880513191223, + "learning_rate": 3.3678440412212576e-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": 0.00434499008115381, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0017850894946604967, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01047506956383586, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.013819062430411578, + "weighted_rkd_loss": 0.00523753478191793 + }, + { + "epoch": 8.017094017094017, + "flow_matching_loss": 0.014346749149262905, + "grad_norm": 0.1989343911409378, + "learning_rate": 3.162006888127833e-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": 0.0036725772311910986, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.001292025181464851, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.008637179713696241, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.014346749149262905, + "weighted_rkd_loss": 0.0043185898568481205 + }, + { + "epoch": 8.034188034188034, + "flow_matching_loss": 0.01425888082012534, + "grad_norm": 0.1949833631515503, + "learning_rate": 2.9625929045422327e-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": 0.003957598004490137, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.001433904911391437, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.009349100943654775, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.01425888082012534, + "weighted_rkd_loss": 0.0046745504718273875 + }, + { + "epoch": 8.051282051282051, + "flow_matching_loss": 0.013923675566911698, + "grad_norm": 0.16496892273426056, + "learning_rate": 2.7696108134781186e-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": 0.00467620249837637, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0017530645069200545, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.011105469428002834, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.013923675566911698, + "weighted_rkd_loss": 0.005552734714001417 + }, + { + "epoch": 8.068376068376068, + "flow_matching_loss": 0.013424544688314199, + "grad_norm": 0.1507738083600998, + "learning_rate": 2.583069056597376e-07, + "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": 0.0038854417391121386, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0017710975953377783, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.009541981061920524, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.013424544688314199, + "weighted_rkd_loss": 0.004770990530960262 + }, + { + "epoch": 8.085470085470085, + "flow_matching_loss": 0.011561490455642342, + "grad_norm": 0.16206304728984833, + "learning_rate": 2.4029757938407116e-07, + "loss": 0.0165, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.004052890441380441, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0017420763499103487, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.009847857290878892, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.011561490455642342, + "weighted_rkd_loss": 0.004923928645439446 + }, + { + "epoch": 8.102564102564102, + "flow_matching_loss": 0.01580977626144886, + "grad_norm": 0.20269455015659332, + "learning_rate": 2.229338903070871e-07, + "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": 0.003932159813120961, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0016555070236790926, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.009519826667383314, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.01580977626144886, + "weighted_rkd_loss": 0.004759913333691657 + }, + { + "epoch": 8.11965811965812, + "flow_matching_loss": 0.013072844501584769, + "grad_norm": 0.19857940077781677, + "learning_rate": 2.0621659797279645e-07, + "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": 0.0038817472057417037, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.001545777020510286, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.009309271443635225, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.013072844501584769, + "weighted_rkd_loss": 0.0046546357218176125 + }, + { + "epoch": 8.136752136752136, + "flow_matching_loss": 0.013049683813005687, + "grad_norm": 0.12958791851997375, + "learning_rate": 1.901464336497233e-07, + "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": 0.00369636001996696, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0015062186168506742, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.008898938726633786, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.013049683813005687, + "weighted_rkd_loss": 0.004449469363316893 + }, + { + "epoch": 8.153846153846153, + "flow_matching_loss": 0.013017881475389004, + "grad_norm": 0.16941314935684204, + "learning_rate": 1.7472410029891527e-07, + "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": 0.0036582418833859265, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.001561993802897632, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.008878477476537227, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.013017881475389004, + "weighted_rkd_loss": 0.004439238738268614 + }, + { + "epoch": 8.17094017094017, + "flow_matching_loss": 0.012620376516133546, + "grad_norm": 0.12690627574920654, + "learning_rate": 1.5995027254319173e-07, + "loss": 0.0167, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.003393899230286479, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0014338911918457598, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.008221689658239485, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.012620376516133546, + "weighted_rkd_loss": 0.004110844829119742 + }, + { + "epoch": 8.188034188034187, + "flow_matching_loss": 0.011777253774926066, + "grad_norm": 0.11281920969486237, + "learning_rate": 1.4582559663764372e-07, + "loss": 0.0164, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.003883430501446128, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0015375308750662953, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.009304391825571657, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.011777253774926066, + "weighted_rkd_loss": 0.004652195912785828 + }, + { + "epoch": 8.205128205128204, + "flow_matching_loss": 0.013854224048554898, + "grad_norm": 0.1607847958803177, + "learning_rate": 1.323506904413485e-07, + "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": 0.0034347342560067775, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0014547378872521223, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.008324206387624145, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.013854224048554898, + "weighted_rkd_loss": 0.004162103193812073 + }, + { + "epoch": 8.222222222222221, + "flow_matching_loss": 0.011663326108828188, + "grad_norm": 0.16224442422389984, + "learning_rate": 1.195261433903594e-07, + "loss": 0.0165, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.004068200034089387, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0015405308920890092, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.009676930774003268, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.011663326108828188, + "weighted_rkd_loss": 0.004838465387001634 + }, + { + "epoch": 8.239316239316238, + "flow_matching_loss": 0.011585106793791055, + "grad_norm": 0.1953214555978775, + "learning_rate": 1.0735251647190969e-07, + "loss": 0.0164, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.004023355850949884, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0015609974274411798, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.009607709199190139, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.011585106793791055, + "weighted_rkd_loss": 0.0048038545995950695 + }, + { + "epoch": 8.256410256410255, + "flow_matching_loss": 0.010779148479923607, + "grad_norm": 0.18219971656799316, + "learning_rate": 9.583034219987408e-08, + "loss": 0.0152, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.003688325686380267, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0014612552186008543, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.00883790673688054, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.010779148479923607, + "weighted_rkd_loss": 0.00441895336844027 + }, + { + "epoch": 8.273504273504274, + "flow_matching_loss": 0.01083608423359692, + "grad_norm": 0.163270503282547, + "learning_rate": 8.496012459148717e-08, + "loss": 0.0152, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.003690268355421722, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0013255010242573916, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.008706037746742367, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.01083608423359692, + "weighted_rkd_loss": 0.004353018873371184 + }, + { + "epoch": 8.290598290598291, + "flow_matching_loss": 0.014668568409979344, + "grad_norm": 0.1465245932340622, + "learning_rate": 7.474233914527618e-08, + "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": 0.0032832271186634897, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0012717477569822223, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.00783820180222392, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.014668568409979344, + "weighted_rkd_loss": 0.00391910090111196 + }, + { + "epoch": 8.307692307692308, + "flow_matching_loss": 0.014427309110760688, + "grad_norm": 0.17697834968566895, + "learning_rate": 6.517743282028254e-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": 0.004399512312375009, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0016026621451601386, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.010401686839759349, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.014427309110760688, + "weighted_rkd_loss": 0.0052008434198796746 + }, + { + "epoch": 8.324786324786325, + "flow_matching_loss": 0.0124134442768991, + "grad_norm": 0.16188935935497284, + "learning_rate": 5.6265824016492604e-08, + "loss": 0.0174, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.004069513431750238, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0019054210744798184, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.010044447798281907, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.0124134442768991, + "weighted_rkd_loss": 0.005022223899140954 + }, + { + "epoch": 8.341880341880342, + "flow_matching_loss": 0.0162894481793046, + "grad_norm": 0.2575010061264038, + "learning_rate": 4.800790255654719e-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": 0.004137730854563415, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0018405434326268732, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.010116005130112172, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.0162894481793046, + "weighted_rkd_loss": 0.005058002565056086 + }, + { + "epoch": 8.35897435897436, + "flow_matching_loss": 0.014363493211567402, + "grad_norm": 0.1337166279554367, + "learning_rate": 4.040402966867862e-08, + "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": 0.004561483091674745, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0017737012123689056, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01089666741900146, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.014363493211567402, + "weighted_rkd_loss": 0.00544833370950073 + }, + { + "epoch": 8.376068376068377, + "flow_matching_loss": 0.012774949660524726, + "grad_norm": 0.1732391119003296, + "learning_rate": 3.3454537970925014e-08, + "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": 0.004006552579812706, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0016211368376389146, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.00963424202054739, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.012774949660524726, + "weighted_rkd_loss": 0.004817121010273695 + }, + { + "epoch": 8.393162393162394, + "flow_matching_loss": 0.015056971274316312, + "grad_norm": 0.14762836694717407, + "learning_rate": 2.715973145656192e-08, + "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": 0.0035704671638086436, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0013850737072061748, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.008526008063927292, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.015056971274316312, + "weighted_rkd_loss": 0.004263004031963646 + }, + { + "epoch": 8.41025641025641, + "flow_matching_loss": 0.01169312335550785, + "grad_norm": 0.1522114872932434, + "learning_rate": 2.151988548082129e-08, + "loss": 0.0162, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.0037407787516713143, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.001625890377908945, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.009107447927817702, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.01169312335550785, + "weighted_rkd_loss": 0.004553723963908851 + }, + { + "epoch": 8.427350427350428, + "flow_matching_loss": 0.012849413696676493, + "grad_norm": 0.14591342210769653, + "learning_rate": 1.6535246748837772e-08, + "loss": 0.0174, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.003902881685644388, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0013816383783705533, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.009187401831150055, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.012849413696676493, + "weighted_rkd_loss": 0.004593700915575027 + }, + { + "epoch": 8.444444444444445, + "flow_matching_loss": 0.013146649487316609, + "grad_norm": 0.15642616152763367, + "learning_rate": 1.2206033304855702e-08, + "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": 0.003651187289506197, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0015399788680952042, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.008842353383079172, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.013146649487316609, + "weighted_rkd_loss": 0.004421176691539586 + }, + { + "epoch": 8.461538461538462, + "flow_matching_loss": 0.013237102329730988, + "grad_norm": 0.17536711692810059, + "learning_rate": 8.532434522695031e-09, + "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": 0.004425908089615405, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.001877948606852442, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.010729764867573977, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.013237102329730988, + "weighted_rkd_loss": 0.005364882433786989 + }, + { + "epoch": 8.478632478632479, + "flow_matching_loss": 0.013877835730090738, + "grad_norm": 0.1576084941625595, + "learning_rate": 5.5146110974679805e-09, + "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": 0.004351154412142932, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.001673866022611037, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.010376174934208393, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.013877835730090738, + "weighted_rkd_loss": 0.0051880874671041965 + }, + { + "epoch": 8.495726495726496, + "flow_matching_loss": 0.012626701407134533, + "grad_norm": 0.19507397711277008, + "learning_rate": 3.1526950385496467e-09, + "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": 0.004268162348307669, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0017047684756107627, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.010241093207150697, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.012626701407134533, + "weighted_rkd_loss": 0.005120546603575349 + }, + { + "epoch": 8.512820512820513, + "flow_matching_loss": 0.010833437461405993, + "grad_norm": 0.20359574258327484, + "learning_rate": 1.4467896637959666e-09, + "loss": 0.0159, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.0043221628991886975, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0015077884832862764, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.010152114275842904, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.010833437461405993, + "weighted_rkd_loss": 0.005076057137921452 + }, + { + "epoch": 8.52991452991453, + "flow_matching_loss": 0.013263375498354435, + "grad_norm": 0.15500546991825104, + "learning_rate": 3.9696959503898733e-10, + "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": 0.003604265209287405, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0013148864556569606, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.008523416798561811, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.013263375498354435, + "weighted_rkd_loss": 0.0042617083992809056 + }, + { + "epoch": 8.547008547008547, + "flow_matching_loss": 0.01256059342995286, + "grad_norm": 0.2574502229690552, + "learning_rate": 3.280754806156416e-12, + "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": 0.004128366347867995, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.001785930647747591, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.010042663337662816, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.01256059342995286, + "weighted_rkd_loss": 0.005021331668831408 + }, + { + "epoch": 8.547008547008547, + "step": 5000, + "total_flos": 0.0, + "train_loss": 0.03792572997510433, + "train_runtime": 9273.3459, + "train_samples_per_second": 34.508, + "train_steps_per_second": 0.539 + } + ], + "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/seed2/default/train_5shot_seed2_fullactionhead_5k/training_args.bin b/FewShot/39tasks/5shot/FromRKD-Timewarp_FullactionHead/seed2/default/train_5shot_seed2_fullactionhead_5k/training_args.bin new file mode 100644 index 0000000000000000000000000000000000000000..03b77c8b5cc649d8c535338805796420e99cc9a9 --- /dev/null +++ b/FewShot/39tasks/5shot/FromRKD-Timewarp_FullactionHead/seed2/default/train_5shot_seed2_fullactionhead_5k/training_args.bin @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:9d4b4312a6d79ef8fa73c97bc91f00d99c6408f7c452df91b898b0233d20c14c +size 6033 diff --git a/RKDv3/base120_single_fm1-rkd0p5_atomic65_human/default/base120_single_fm1-rkd0p5_atomic65_human.yaml b/RKDv3/base120_single_fm1-rkd0p5_atomic65_human/default/base120_single_fm1-rkd0p5_atomic65_human.yaml new file mode 100644 index 0000000000000000000000000000000000000000..06a36f90d5c783ef2c99df00ce7ee0405506fa8a --- /dev/null +++ b/RKDv3/base120_single_fm1-rkd0p5_atomic65_human/default/base120_single_fm1-rkd0p5_atomic65_human.yaml @@ -0,0 +1,36 @@ +experiment_name: RKDv3 TimewarpVAE DA, Student:APHead, single-phase FM+RKD 1:0.5 from + GR00T base to 120k on Atomic65 human +policy_type: groot_RKDv3 +base_model_path: nvidia/GR00T-N1.5-3B +output_root: ${HOME}/groot_robocasa/robocasa_v2/RKDv3/base120_single_fm1-rkd0p5_atomic65_human +dataset: + dataset_soup: atomic_pretrain_human +training: + num_gpus: 2 + batch_size: 64 + seed: 42 +model: + rkd_teacher_source: timewarp_action + rkd_timewarp_vae_checkpoint: ${HOME}/groot_robocasa/TimeWarping/TimewarpVAE/robocasa_vanilla_t16_z8_g8_h32_reg0p01_beta0p01/checkpoint-step-60000-epoch-17.46 + rkd_relation_mode: flatten + rkd_loss_type: distance_angle + rkd_student_reconstruction_head_pooling: attention + rkd_student_reconstruction_head_enabled: true + rkd_student_reconstruction_head_dim: 512 + rkd_student_reconstruction_head_hidden_dim: 512 + rkd_distance_loss_weight: 1.0 + rkd_angle_loss_weight: 2.0 + rkd_exclude_diagonal: true +phases: +- name: fm_rkdv3_timewarp_vae_da_aphead_fixed_0p5_120k + max_steps: 120000 + save_steps: 60000 + 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 diff --git a/RKDv3/base120_single_fm1-rkd0p5_atomic65_human/default/fm_rkdv3_timewarp_vae_da_aphead_fixed_0p5_120k/experiment_cfg/metadata.json b/RKDv3/base120_single_fm1-rkd0p5_atomic65_human/default/fm_rkdv3_timewarp_vae_da_aphead_fixed_0p5_120k/experiment_cfg/metadata.json new file mode 100644 index 0000000000000000000000000000000000000000..c97a2242b0aa3efb4b5182154559d4db7cced526 --- /dev/null +++ b/RKDv3/base120_single_fm1-rkd0p5_atomic65_human/default/fm_rkdv3_timewarp_vae_da_aphead_fixed_0p5_120k/experiment_cfg/metadata.json @@ -0,0 +1,431 @@ +{ + "new_embodiment": { + "statistics": { + "state": { + "base_position": { + "max": [ + 7.778976917266846, + 0.4876587688922882, + 0.7223060727119446 + ], + "min": [ + -4.94293737411499, + -6.890198230743408, + 0.6994727253913879 + ], + "mean": [ + 2.645730005800653, + -2.110066447108802, + 0.7004254938507055 + ], + "std": [ + 1.674798582069234, + 1.350392482318102, + 0.001107956680671985 + ], + "q01": [ + -1.0132281893317567, + -5.366714195503661, + 0.699973412653353 + ], + "q99": [ + 6.099414437993715, + -0.5498886012762193, + 0.7032860957123939 + ] + }, + "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, + 0.24103797327895668, + 0.6071393324239354 + ], + "std": [ + 0.0, + 0.0, + 0.6615962784298529, + 0.36814045893518177 + ], + "q01": [ + 0.0, + 0.0, + -1.0, + 3.3913522062507517e-06 + ], + "q99": [ + 0.0, + 0.0, + 1.0, + 1.0 + ] + }, + "end_effector_position_relative": { + "max": [ + 0.9014528393745422, + 0.8569459915161133, + 1.053884744644165 + ], + "min": [ + -0.5828549861907959, + -0.8486236333847046, + -0.3446122705936432 + ], + "mean": [ + 0.2641355221895152, + -0.01768191244102862, + 0.4458847037294584 + ], + "std": [ + 0.16437879912713016, + 0.24835923712456218, + 0.2124190803465676 + ], + "q01": [ + -0.004962867024380738, + -0.46862868595225443, + 0.20316236863116172 + ], + "q99": [ + 0.5495510296644974, + 0.44201455499146414, + 0.6984316542098391 + ] + }, + "end_effector_rotation_relative": { + "max": [ + 0.9999998807907104, + 0.9997354745864868, + 0.9900747537612915, + 0.9735639095306396 + ], + "min": [ + -0.9999930262565613, + -0.9999884366989136, + -0.9975369572639465, + 6.754255821306288e-08 + ], + "mean": [ + -0.23536101367097556, + 0.0384247508727208, + -0.06854891687606551, + 0.1400254132624415 + ], + "std": [ + 0.8296216819146821, + 0.30067208350163366, + 0.33390768096919515, + 0.16926885514344236 + ], + "q01": [ + -0.9950540525525874, + -0.5590028004550295, + -0.5635208777456633, + 0.001747048046951923 + ], + "q99": [ + 0.9472459325196646, + 0.6183763159959191, + 0.5108060516185887, + 0.46646525101830405 + ] + }, + "gripper_qpos": { + "max": [ + 0.055869169533252716, + 0.010916369967162609 + ], + "min": [ + -0.011669117026031017, + -0.06307756155729294 + ], + "mean": [ + 0.031849369408926076, + -0.03200748595341724 + ], + "std": [ + 0.012248040928874927, + 0.01206276030599065 + ], + "q01": [ + 0.006329614505051435, + -0.0405872199089872 + ], + "q99": [ + 0.040445365520021676, + -0.006282430560590391 + ] + } + }, + "action": { + "base_motion": { + "max": [ + 1.0, + 1.0, + 1.0, + 0.0 + ], + "min": [ + -1.0, + -1.0, + -1.0, + 0.0 + ], + "mean": [ + 0.0071452133647496785, + 0.0015603604277540566, + -0.0014052511062474704, + 0.0 + ], + "std": [ + 0.11471999232420287, + 0.10209228015027566, + 0.0855718789814838, + 0.0 + ], + "q01": [ + -0.13329522793794643, + -0.16270717101832055, + -0.12619798329427676, + 0.0 + ], + "q99": [ + 0.1721366453818688, + 0.1972873452547982, + 0.11128571706089413, + 0.0 + ] + }, + "control_mode": { + "max": [ + 1.0 + ], + "min": [ + -1.0 + ], + "mean": [ + -0.9218734936549906 + ], + "std": [ + 0.3874975703570591 + ], + "q01": [ + -1.0 + ], + "q99": [ + -0.26692269247002653 + ] + }, + "end_effector_position": { + "max": [ + 1.0, + 1.0, + 1.0 + ], + "min": [ + -1.0, + -1.0, + -1.0 + ], + "mean": [ + 0.002920238070268553, + -0.0060913649045883315, + -0.07322447393298565 + ], + "std": [ + 0.4482114094812734, + 0.41233706159751987, + 0.4006084454909327 + ], + "q01": [ + -0.934104737554819, + -0.9255792016234818, + -0.8085822192859831 + ], + "q99": [ + 0.7874185230421324, + 0.9004049161564378, + 0.8018332495119154 + ] + }, + "end_effector_rotation": { + "max": [ + 1.0, + 1.0, + 1.0 + ], + "min": [ + -1.0, + -1.0, + -1.0 + ], + "mean": [ + 0.0077772595328458915, + -0.022620543966259083, + -0.00029332176634943376 + ], + "std": [ + 0.11155576533045976, + 0.13202104364511658, + 0.1240098754411023 + ], + "q01": [ + -0.2726786325766535, + -0.4125887853047325, + -0.3445183934868137 + ], + "q99": [ + 0.31803481867756295, + 0.3015153519871841, + 0.34730986869917346 + ] + }, + "gripper_close": { + "max": [ + 1.0 + ], + "min": [ + -1.0 + ], + "mean": [ + -0.33422761717726496 + ], + "std": [ + 0.9425001619711008 + ], + "q01": [ + -1.0 + ], + "q99": [ + 0.681821585371573 + ] + } + } + }, + "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/RKDv3/base120_single_fm1-rkd0p5_atomic65_human/default/resolved_config.yaml b/RKDv3/base120_single_fm1-rkd0p5_atomic65_human/default/resolved_config.yaml new file mode 100644 index 0000000000000000000000000000000000000000..115ecea22254bc15d71c1edaeacda693cd6954e1 --- /dev/null +++ b/RKDv3/base120_single_fm1-rkd0p5_atomic65_human/default/resolved_config.yaml @@ -0,0 +1,37 @@ +experiment_name: RKDv3 TimewarpVAE DA, Student:APHead, single-phase FM+RKD 1:0.5 from + GR00T base to 120k on Atomic65 human +policy_type: groot_RKDv3 +base_model_path: nvidia/GR00T-N1.5-3B +output_root: ${HOME}/groot_robocasa/robocasa_v2/RKDv3/base120_single_fm1-rkd0p5_atomic65_human +dataset: + dataset_soup: atomic_pretrain_human +training: + num_gpus: 2 + batch_size: 64 + seed: 42 +model: + rkd_teacher_source: timewarp_action + rkd_timewarp_vae_checkpoint: ${HOME}/groot_robocasa/TimeWarping/TimewarpVAE/robocasa_vanilla_t16_z8_g8_h32_reg0p01_beta0p01/checkpoint-step-60000-epoch-17.46 + rkd_relation_mode: flatten + rkd_loss_type: distance_angle + rkd_student_reconstruction_head_pooling: attention + rkd_student_reconstruction_head_enabled: true + rkd_student_reconstruction_head_dim: 512 + rkd_student_reconstruction_head_hidden_dim: 512 + rkd_distance_loss_weight: 1.0 + rkd_angle_loss_weight: 2.0 + rkd_exclude_diagonal: true +phases: +- name: fm_rkdv3_timewarp_vae_da_aphead_fixed_0p5_120k + max_steps: 120000 + save_steps: 60000 + 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 +resolved_sweep: {}